I just finished updating Genunix and all its associated mirrors. There have been recent releases of all 3 unofficial distros and you can get them all, including the latest official source, from Genunix.

One word of caution for nVidia users. OpenSolaris (meaning, the Solaris Express: Community Release and the unofficial distros) will bomb out when starting X if you've got a PCI Express nVidia graphics card. The screen will just look mangled and your console will likely become completely unresponsive requiring you to do a hard reset on the box. This is because you must run the nVidia official driver (download it here). To get around this, install the system from the console without X, then, when fully installed, download and install the nVidia driver and restart the system. With the nVidia driver installed the system will work perfectly (in fact, better than perfect, the driver is really nice). So if you've been frustrated by OpenSolaris on your box and can't figure out why things are going nutty, thats what you've got to do. I've tested Nexenta's LiveCD and it suffers the same problem although I haven't personally tested BeleniX and SchilliX yet.

Talk about kool, after a lot of hard work the buys at Greenplumb have released Bizgres MPP 2.1. You can download it from the new 'Bizgres Network'. And these guys are serious about this release, offering support for trial users for 30 days, via both phone and email.

What could this release possibily mean to IBM DB2 and Oracle 10g RAC? MPP is a place where DB2 has thrived for years and now has a very interesting competitor, and an open source one at that. I only wish I had a couple more machines to really put this thing through its paces. :(

Congrats to GreenPlum on a job well done. Keey your eyes on these guys!

Efficiency. Performance. Value. Enjoyment. Geeks gotta have it all, and we're busy, so we don't have a lot of time. Here are 4 things that I think every geeks gotta love for the home.

1. Weber Q Portable Propane Gas Grill
   

   I bought this grill almost a year ago and its been the greatest addition to our cookware in years. We eat a lot of chicken because its so easy to work with, but after a couple years of baked chicken you need a break. With the addition of this grill we greatly increased the range of protien sources in our diet. Because of the kids we can't go out to steak houses anymore, but meat is pretty cheap, so we now eat better steaks and ribs at home than we ever did out on the town, and at a faction of the cost. The grill is easy to work with and to clean and most importantly has a real, solid, cast-iron grill plate. You can grill up almost anything on this baby and it doesn't take up a lot of space.

   Prior to this we had an old standard Weber charcoal grill, but it takes like 30-40 minutes to get prep'ed and its difficult to cook on because of the high heat, plus the cooking grate sucks. We still use the charcoal when we want to sear ribs, but thats about it.

2. Mr Clean AutoDry Car Wash System Starter Kit
   

   I got tired of washing my bad ass Volvo S70 T5 at my local joint and playing $20 a pop. They did a lack-luster job and it took time out of the day. So I saw this thing and thought that at $20 it was at least worth a shot. Gotta say, I love it. I can wash the car completely in about 5 minutes with it, and it dries quickly completely spot free. The first time you use it there is some adjustment to how things work, but once you get the hang of it you can really fly. I've even washed the car while waiting for Tamarah to finish getting ready to leave.

   The only possible downside to the product is that while the car looks beautiful when your done, the car isn't waxed, so you build up dirt and dust more quickly that you would with a proper wax on it and I've not yet found a waxing system thats as fast and flawless as this washer is.

3. Sony DAV-FX100W Wireless Home Theater Dream System
   

   I bought one of the Sony "all in one" DVD home theaters when they first released in the US, and I still use it to this day. Since then the number of options and goodies has grown and there are several to choose from, both from Sony and others. But I still like the Sony's. This particular model is nice because it eliminates one of the great shortfalls of the all in one solution. The whole point of having a compact solution like these is to give you everything you want, but in very little space, so that apartment dwellers and people with homes that aren't suited for full home theaters can be happy, but all that gets kinda messy when you start stringing speaker cable all over the place. Since in an apparment or rental house you can't just rip up all the carpet you have to get creative and in some cases that means just settling for 3.1 sound by nix'ing the rear channels. But now you can get it wireless! w00t. When my DVD player bites the dust eventually this is definately the way I'll go.

4. Sun Ray 1
   

   Finally, the essential home geek utility, the Sun Ray 1. You just can't beat these little guys. They use almost no power at all, cost less than $50 on eBay (frankly, if you pay over $20 your getting ripped) and the software is free. Take your noisy PC, whether its got Windows, Linux, or Solaris, and shove it in your closet (well ventalated closet of course) and just leave this baby on your desk. Noise? Whats that? And they are so flexable that you can put them in any part of the house that you can run Ethernet to. I'm thinking about putting one in the bathroom! With your smart card you can just plug in, alhough you don't need 'em, just login in manually on the keyboard and your set to go. No matter who you are or what your doing, you can use a Sun Ray to get it done. These guys are so massively underappreciated its not even funny. If you need even more fun, put the thing into Kiosk mode and bury it in the wall to make a handy home automation interface. These guys just kick ass.

There you have it, my list of handy gadgets to have around the geek house. Affordable and efficient, and in my case, proven. Why do y ou care about these? No idea, but sometimes we all need a break from boring and technical blogging and toy around with something else to mix it up. :)

You might have seen TV ads or billboards around town several months ago featuring various ladies advertising Dove's Campaign For Real Beauty. The ads features a following gaggle of gals:

I'm very glad to say that this wasn't a short term Dove marketing scheme but in fact a long term commitment to the ideal. And frankly, I love the idea. I've been a big supporter of the effort from the beginning and glad to see it continue to gain steam and even spread around the globe. The media has given our girls a false sense of beauty, one that is marketable because its so unusual, unattainable, and unhealthy. Working our way out of that rut is a long and difficult battle thats fought one girl at a time.

To this end, Dove has put together The Dove Self-Esteem Fund. Girls can sign up to recieve workbooks and information that can help change their perceptions and encourage them to embrace who and what they really are. And a girl who cares for herself, takes care of herself, and lives a happier, fuller, more beautiful life as a result. Most of us know from experience that "beautiful people" can look pretty unattractive when they open their mouths and spew shallow bs, and that perhaps less convensionally attractive people can start looking more and more beautiful because of who they are and the smile on their face. And even more basic a fact is this, for every man there is a woman, and for every woman a man... simply math tells us that there is no such thing as an unattractive woman or man, just people who haven't found eachother yet.

So if you know a young girl, or even an older stateswoman, who's having self esteem issues, turn them on to the Campaign For Real Beauty. Encourage them to dig around, view at the TV spots, read through the profiles and rate other real women on the site. You can even register for a free shirt... although, its free with $15 worth of Dove products, but I'll tell you that Dove Body Wash is amazing stuff; Tamarah switched from Dial soap to Dove a while back and her skin feels like creame. Its amazing stuff. Who better to know beauty than Dove. :)

Here we are, February slipping away, and still Oracle has not released Oracle 10g R2 for Solaris/X86. I am really frustrated by this. I've been hounding Oracle about this since October, and every time I ask I'm assured that the release is just weeks away. The announced renewal of collaboration in January gave me some hope that finally we'd get things back on track and I'd shortly see the release of R2, Express, and Instant Client for Solaris10/X86... but 2 months later there is still nothing to show. Are the boys at Oracle working feverishly? Possibly, but I'm still waiting and my fingers are aching from the daily Oracle Download page refreshes.

Ingres and Sybase are both said to be working toward Solaris 10/X86 releases, but they aren't ready yet. Some of the frustration of platform support is caused by old terminology, that is, that "Solaris 64bit" just automatically means "SPARC". In the case of Sybase ASE this recently hit me when I downloaded Sybase ASE 15 Developers Edition, Solaris 64bit, which since it was lumped together with Linux and Windows, I assumed wrongly that this would be X86... in fact its supported on Solaris 8/9/10 SPARC. DB2 is the same way, although, odd as it seems, Informix is avalible for Solaris/X86.... its always the one that you don't want. ;)

Now, I realize that the database vendors are busy and up against constantly shortening deadlines and that having a new platform thrown at you can be a real curve ball, but given that they've already got the products running on Solaris/SPARC the bulk of the work is done... its certainly not so simple as a quick reguild and ship, but its not like porting to BeOS or something.

So, how hard are these vendors working on Solaris/X86 support? Is Sun pushing on these vendors? Are these vendors being pushed by customers? I'm keeping my voice in Oracle's ear as best I can, but perhaps there needs to be more outcry to IBM for DB2 and Sybase for ASE. What might be even more frustrating to them is that as soon as they finish it we'll expect it for free. But, the upshot is that with Sun's new Galaxy X64 server line the likely hood of seeing larger Solaris installations leveraging technologies like Oracle 10g RAC and DB2 DPF are much more likely providing a real oppertunity for sales. When you get customers deploying on X4200 systems, your scalability is limitless, whereas single system deployments on Sun midrange is going to hit a limitation when you max out the box, untill you scale into a 2 box setup which is costly.

Don't get me wrong, I'm a great lover and fan of the Sun Midrage SPARC systems, especially when deployed with Sun UltraSPARC IV+ processors, but SPARC and X86 solve diffrent problems, and similarly, Solaris X86 and Solaris SPARC are integral parts of seperate solutions. For an HR database that needs to be highly reliable I'd tend toward Sun Fire E2900 with Sun Cluster. But for a large web facing database where HA failover isn't good enough, I'd want a faster and more flexable solution, such as Oracle 10g RAC or DB2 DPF on Sun X4200 servers. The only outstanding questions really is, how well are these databases going to perform on Niagra? And thats what I'm waiting to see...

In the mean time.... wait wait, I'll just keep waiting.

I know, when you think of exciting video codecs, Real Player is just right up there at the top of your list. But, believe it or not, the Real Player codecs really have come a long long way, the only shame is that so few people are using the latest codec. Having Real Player in Solaris/X86 following Build 31 is a great advantage, not the least of which is that you can now actually watch Sun released videos on Solaris (which is a sorta nice thing). But there is more to it than that. If you haven't noticed, Real Player 10 included supports the Xiph Ogg codecs including OggVorbis and OggTheora!.

This is majorly frickin' kool. I'm a huge believer in Xiph. You might recall that I wrote Eplayer, an OggVorbis audio player based on the Enlightenment Foundation Libraries. Xiph's codecs are beautiful peices of art; easy to use and very flexable. Using Xiph's no defunct libao I had a solid cross-platform foundation for audio handling, and libogg and libvorbis are so easy to use that I was spending my time improving my user interface, not sound. The entire suite of codecs are first class in my opion. If you don't already know, Ogg is itself only a container, similar to Quicktime MOV. Various codecs, such as Vorbis and Theora utilize that standard container format simplifying the way you interact with the data. Xiph's suite of codecs include:

• Vorbis: Audio
• FLAC: Losses Audio
• Speex: Speech optimized Audio (for use with VoIP, etc)
• Theora: Video

Real Player 10 includes support for Vorbis and Theora. That means that its time for me to start using Ogg Theora widely again. I just love the codec and love having a solid alternative.

If you have Real Player 10 (Solaris Express: Community Release B33) and want to try out some Theora for yourself, here is a OggTheora highly compressed version of the first Silicon Valley OpenSolaris Users Group meeting.

If your interested in Real Player, you can help work on it yourself by visiting the Helix Community.

During the day I'm the SA of a small-ish hosting provider. One of the key peices of our enviroment is our load balancers. For years we've been using old F5 BigIP load balancers to handle traffic. Now its time to find a new solution, and so about 3 months ago we started doing evals and seriously looking at options. Normally I don't handle front end networking gear, but because just about all load balancers run some varient of UNIX I end up doing the bulk of the testing and work on the units. When we started this process the obvious choice was to look at the new F5 offerings. However, I wanted an alternative, so, having heard good things, I decided to bring in NetScaler, who about that time got gobbled up by Citrix. I also considered Foundry ServerIron, but seeing that it was really just a Foundry switch with some extra software I quickly dismissed it.

I say "load balancing", but the current lingo is actually "application switching". Some people even say "content switch" or "layer 4-7 switching". The idea is the same whatever name you give it, make a bunch of things look like one. That is, present a Virtual IP ("vip") to the outside world, which actually sends those incoming requests to one or more backend servers by some method, such as round-robin. This way you can serve more content than any one system could. Things get more interesting when you add "content switching", meaning that you pick data out of the actual content itself (typically an HTTP header) and make decisions about where it should go. Using content switching functionality in a load balancer you could, for instance, send all HTML requests to one set of servers and all image requests to another set of servers, with the end user never knowing whats actually happening on the backend.

Load Balancing isn't cheap and its a critical part of the infrastructure, so we needed to carefully choose. So we've done both in house testing of both the F5 BigIP and NetScaler. Here are some notes..

NetScaler

The NetScaler actually runs BSD under the hood. When you SSH to it you are dropped directly into an IOS-like shell for management of the device. Because of this, the device feels and acts like a network device. tab-tab your way around the interface to find what you want. The CLI is, therefore, extremely easy to use. You can also print out the running config (show runningconfig) at any time, which you can then cut and paste off to a file quickly without bundling flles and moving tarballs, making it easy to do quick config snapshots while working on a complex configuration.

To actually create a VIP you first add a bunch of servers, then you add services which details the services provided by the servers such as which protocal (HTTP), which port (80), client and server timeouts, and other related tunable parameters for that service. Once thats in place you can create an actual virtual server with an IP that your end-users will see, to which you then bind one or more of the services to that virtual server.

> add server ironman 192.168.1.25
 Done
> add server mystique 192.168.1.83
 Done
> add server thor 192.168.1.73
 Done
> add service ironman_http ironman HTTP 80
 Done
> add service mystique_http mystique HTTP 80
 Done
> add service thor_http thor HTTP 80
 Done
> add lb vserver Test-Vserver-80 HTTP 192.168.10.100 80
 Done
> bind lb vserver  Test-Vserver-80ironman_http -weight 1
 Done
> bind lb vserver  Test-Vserver-80mystique_http -weight 1
 Done
> bind lb vserver  Test-Vserver-80 thor_http -weight 1
 Done

There is a little more to it than this, but not much. The NetScaler is easy to work with, direct, and does what you expect a load balancer to to.

Extra functionality, like content caching, SSL offload, persistant connetions, and content switching are also easy to add. The only bit thats funky is content switching, where you are forced to create a whole new content switching virtual server (vserver), which is lame because I'd much rather create content switching rules and apply them to existing virtual servers. This means that if we want to create new content rules or modify existing ones we can't just apply it to an existing virtual server and then yank it seconds later if things aren't working properly. The actual rules themselves though are easy to create and look similar to pattern matching that you'd see in PERL or elsewhere.

So, the big plus is that it looks, feels, and smells like an actual network device with an isolated shell and the works. Its easy to work with, provides excellent performance, and does just about everything you generally expect. To put it another way: it is what it is, nothing special or exceptional, but it does the job well.

F5 BigIP

When we considered buying newer F5 units I was already holding 2 strikes against them. As a long time cusotmer I'd put up with at least 2 complete OS changes on the units (both were presented as patches that installed without incident, but it was annoying nevertheless), a painful CLI interface (bigpipe), and a web interface that made you want to put up with the horrible CLI. Even more annoying was the fact that they have a tendancy to modify the configuration formats that BIgIP uses, which means that almost every upgrade invalidated all of our prior configuration backups and meant re-learning the configuration layout.

But, it would be stupid not to consider F5, who with 33%, is the market leader. By contract, NetScaler is #5 on the list. So they agreed to give us two high end units for 2 weeks of testing, although they required that we fill out these forms about our testing, experiences, and results, which made me less than excited. The units arrived, we had a sales meeting, and the sales engineer (Nathan) stayed around for the afternoon to help get me up to speed as quickly as possible. The one advantage I had with the BigIP was that I had amassed several years of experience with the older products and pretty well knew how things worked according to F5.

On BigIP you construct VIPs alittle diffrently. First you configure a bunch of nodes, including all node specifics. Then, unlike NetScaler, you group one or more of these nodes into a pool. Then you create a virtual server with its IP and port, and associate one or more of the pools with that VIP. Now, the distinct advantage here is that you can swap out all the backend servers behind a VIP quickly, rather than unbinding and rebinding services like on the NetScaler. One of the ways that we use pools here is to create "primary" and "secondary" pools of servers. When we do upgrades or site changes or things we can "switch to secondaries" by simply moving the pools, and when we're done, switch back just as easily with a single action. The abstraction that pools provide really simplifies management and has several other advantages.

One of the feaures of BigIP 9.x that F5 is really proud of is called iRules. IRules are TCL code snippets which can modify the way that your VIP handles traffic. The most comment example is to redirect requests to diffrent pools within a virtual server based on some criteria, such as if the URI contains .jpg, redirect to the IMG_Pool, otherwise continue as usual. iRules can get much more interesting, however, because you effectively have the packet headers open in your hands while that iRules is executing, meaning that you can slip new information into the header as it passes by for special handling by your web servers, or to modify content. One example of the power of iRules that F5 is proud of is a customer who wrote an iRules to modify credit card numbers, not only intercepting them, but validating that they are (by means of the common algorithm) actual valid numbers, and then X'ing them out as they are passed to the user. Now, this is something that should be done by the back-end application, however using an iRule they could catch any possible slip-ups and never allow a credit card number to be returned to the customer in the clear. iRules get used by BigIP in other places under the covers, such as by "stream profiles", which are effectively a search-and-replace function, so that any instance of "cuddletech.jpg" could be replaced with "p0wned.gif" without having to modify user content. Again, something that should be done on the servers themselves, but having the ability to do these types of things on the load balancer means that I can put double-checks in place and also compensate for an application flaw temporarily while the back-end application or content is fixed.

Several improvements have been made to the BigIP clustering functionality. BigIP's can be run either Active-Active or Active-Passive, although no one recommends that you use Active-Active. One of the problems with failover in the past was that you need to ensure your configuration is consistant between the two units before you do it, which is done by means of a ConfigSync. But there was no way to really know that the configurations were in proper sync without digging into the config files themselves. In the 9.x releases there is a new feature: ConfigSync Monitor which keeps track of consistancy between the units and goes even further by putting a "ConfigSync Status:" inidcator at the top of the web interface. Thats a big confidence builder when preparing to failover the boxes. Otherwise HA on the units works more of less the way it always did before, with some extra backend goodies, but nothing mind blowing.

If you used an old BigIP you might have wondered about the old RS-232 serial cables that were used between boxes. This failover cable seemingly was a heartbeat or something, but as it turns out no data exchange happens on that cable, its a simple electrical indicator that the other box is running, and if the voltage on the line disappears, obviously the other box blewup. Well, that cable is still there, although you no longer need it. Communication now can be done soley by means of talking on the internal and, optionally, the external, vlan's.

Conclusion

NetScaler is the underdog, the little guys that are trying hard, and I like that.... but sadly, when your stuck with something for 5 years you have to look at the technology itself reguardless of whether you like the company or not. The NetScaler unit is a wonderful network device and if I needed a small load balancer that I didn't want to touch, I'd definately consider the NetScaler. But when it comes to managing a large site thats dynamic and may require situations in which we bring up half the site and drop the other half, while keeping all that activity hidden from the end user, the F5 just comes out on top.

When it comes to high-end functionality and not only routing packets based on content but also having the ability to stop, think, and even change that data during the exchange is really compelling. F5 prides itself on running two independant TCP/IP stacks and on having the superior performance to pull that off, and because of it, they can leverage a huge amount of functionality. When looking seriously at a complete solution its hard to ignore that capability and the potential ways you could leverage it today and tomorow.

So, if you couldn't already tell, we're choosing F5. I hate to say it, but we're stickin' with the big guys. The product just really stands on its own and offers a wide range of new possibilities that we just can't ignore. In the end, we didn't choose on peformance, or even HA capabilities, but on the ability to truly have fine grain control over whats coming in and going out.

So Google is digitizing the American National Archives. I think thats fantastic. Our history is dished out slowly inbetween commercials on the History Channel or availble by making a migration to Washington DC. Putting everything online, video at least, is a wonderful move and I'm excited by it. Watching the moon landing was really kool. I can't wait to see more.

Similarly, if you haven't ever visited the website of the US National Archives, I highly suggest that you do. A large number of US Historical documents are avalible, scanned in at enormous resolutions, allowing you to see ever scratch on the paper and ever detail of the signatures. I find looking over the Constitution to be a moving experience, personally.

On a less serious note, here are two good friday funnies for you:

• This one is safe for work.
• This one is NOT safe for work.

I mentioned my SDN interview about 2 weeks ago, and now its up online. View the SDN interviews here.

This was filmed in "one take" just after 11am, and my brain doesn't come online till around noon, so I'm not exactly verbose. I wish there was some real content in it that would be useful to folks instead of just a marketing piece, but thats what they wanted I guess. The fact that my interview is second behind Big Boss Jonathan Schwatz is kool. So the big up shot is that this is the first SDN monthly that is 66% long hair dominated. ;)

For any rugby fans who are wondering what jersey I'm wearing, its Auckland NPC Rugby Union jersey. I'm 100% born and breed American, but I do love Rugby Union, and I figured being the only person in the webcasts wearing a Sun shirt (which I had underneight) was a bad idea.

As a funny aside, I never touched that foosball table, nor have I ever played. Real men don't play with dolls on sticks and ping pong balls... they scrum, like God intended.

Its that time again, hard to believe that its already the end of Feb! Damn. The upshot is that tonight at the Feb meeting of the Silicon Valley OpenSolaris Users Group we'll be learning about something near to my heart: SATA Frameworks! Storage geeks unite, tonight should be a fun and informative meeting. I can't wait. See ya there!

For the rest of you who can't attend, I will be taping this meeting, so I'll have the video up hopefully over the weekend.

The February edition of NetApp's Tech OnTap newletter is up. If your interested in NetApp Filers or are considering them, give it a look. It sucks that you have to register for it, but you know how all these corps have to show metrics on user statistics in order to keep projects alive, so help someone keep their job by registering for it. Some goodies in there. In particular I found the SnapMirror® Best Practices Guide and iSCSI Multipathing Possibilities articles interesting. By leveraging NetApp Filers with Solaris 10 Update 1 you can do some amazing things, no other combination of system and storage really opens up the possibilities quite so much and provides so much value. Especially these days when applications expect more and more that the shared storage medium just handles redundancy and data protection, these technologies in OnTap get more and more interesting.

In related news, NetApp is trying to get me a spare unit for testing and evaluation. If I get that box I'll be able to share with you guys all the interesting things that OnTap 7g (the newest NetApp Filer OS) can really do. Filers are about more than just simple NAS, in fact, I still firmly believe that Ethernet NAS is stupid, but reguardless, the functionality that OnTap brings to the table is unique and compelling enough that even hating storage over ethernet is a minor concern compared to the new possibilities. Hopefully I can soon share some examples of what I mean.

So we've got 2 videos pertaining to OpenSolaris on Google Video. One by me and one of me. We need more... lots more.

I need feedback as to whether or not the SVOSUG Video looks good on Google. If folks like it I'll stop posting the divx video (found on the SVOSUG page) and start posting it all to Google.

The chief advantage of using Google is that you don't have to download the video, waste disk space to hold it, or worry about installing a suitable video player. I think its the way to go, although they drop the framerate slightly which sucks, but its a fair tradeoff in my opinion.

If you think its a good idea, say so, and we'll start transitioning everything over.

Download DB2 Univeral Database Enterprise Server Edition v8.2. File name? DB2_V82_ESE_SUN_3264_NLV.tar. Looks like v8.2. Says so on the download page. So, download it, untar it, and start to install it. What do you get?

benr@anysystem ~$ cd /opt/IBM/db2/V8.1/bin
benr@anysystem bin$ pkginfo | grep -i db2
application db2ca81                          Configuration Assistant
application db2cc81                          Control Center
application db2chen81                        Control Center Help (HTML) - en_US.iso88591
application db2cj81                          Java Common files
...
benr@anysystem bin$ pkginfo -l db2engn81
   PKGINST:  db2engn81
      NAME:  Base DB2 UDB Support
  CATEGORY:  application
      ARCH:  sparc
   VERSION:  8.1.0.64
   BASEDIR:  /opt
    VENDOR:  IBM Corporation
      DESC:  Base DB2 UDB Support for Solaris
    PSTAMP:  20040813
  INSTDATE:  Feb 22 2006 02:47
   HOTLINE:  Please call IBM Software Support centre
    STATUS:  completely installed
     FILES:      341 installed pathnames
                  22 shared pathnames
                  40 directories
                 182 executables
              317130 blocks used (approx)

Don't know about you, but looks like v8.1 to me. Most of the commands in DB2 are dependant on Java and everything complains in a way that makes no sense at all:

benr@anysystem bin$ ./db2fs
DB2JAVIT : RC = 9501
DB2INSTANCE : -2029059916

Very informative indeed. In fact, this is where I left off with DB2 about a year ago when I tried it... I never got things to work. Most of the Linux users that had these types of problems upgraded to 8.2 and everything was fixed, so I'm assuming that even on Solaris/SPARC that might be a good course of action, but how can I run 8.2 when supposedly this IS 8.2.... except its not, its 8.1.

Why do I hate DB2? Because I can even use it, and when I try, it turns out that IBM can't even be bothered to put the right release in the right package. This eval has been on the site for over a year, I mean, has no one reported it in all that time or do they really just not care? For what its worth, I think I'd really like DB2 if I could get it to run, but not today. I've decided to request that a sales guy call me and at least I can maybe get this straightened out. Way to go IBM.

UPDATE (24 hours later):

So I'm quickly warming up to DB2. Here's the run down...

1. I was bugged by the fact that I downloaded DB2 UDB v8.2 but the installed version was v8.1. Well, it turns out that in fact I am running 8.2. The versioning of DB2 is currently really funky, this is how it works out:

    V8.1 FixPak 1-6 == 8.1
    V8.1 FixPak 7   == 8.2
    V8.1 FixPak 8-11 == 8.2 FixPak 1-4
   

   So if you look in the Release Notes (/opt/IBM/db2/V8.1/Readme/en_US.iso88591/Release.Notes) you'll see that the downloaded trial is DB2 v8.1 FixPak 7, which is DB2 v8.2. So the versioning is really decieving untill you figure out how the versioning works. In short, this concern was a non-issue in the end, but I can't imagine why IBM chose to do this.

2. The second issue I had was also a non-issue but something I reported to IBM. There are two installers in the tarball you download: db2setup which is a GUI installer, and db2install which is a CLI installer. When installing with db2install it gets most of the way through the package installs and then dies due to a bad argument to test. IBM is aware of this issue and it will be "fixed in a future release".

3. The third and actually impacting issue was the errors I was getting from all the utilities. I couldn't run anything. The errors looked like this:

   $ db2fs
   DB2JAVIT : RC = 9501
   DB2INSTANCE : -2029059916
   

   and...

   $ db2
   SQL10007N Message "-1390" could not be retrieved.  Reason code: "3".
   

   IBM thankfully provides trial support via a web interface. I put in a request today about all of these issues and a fellow (my main man Rudy) got back to me about 2 hours later. He was friendly, sympathetic, and extremely helpful. He answered the first two problems, but not this one about the utilities not working but asked for some information to research the issue. Isn't that nice of him? No cost trial support and the guys doing research for me! I'm very very impressed. Go Rudy!

   So Rudy pointed me to the latest DB2 FixPak (FixPak 11, aka DB2 v8.2 FixPak 4) but pointed out that he didn't think it would solve my problems, but try it anyway. Sounded like a good idea either way, so I downloaded (640MB!!!) and installed it. The FixPak installation is really smooth and simple, just unpack and run the included install script. Sure enough, after it completed, nothing had changed for me. So I kept looking around with the help of Google.

   In the process of researching these problems I've flipped through hundreds of pages of documentation, and honestly the more I learn about DB2 the more I'm interested in it. Its definately not a dog and its definately got some kick ass features. There are obviously problems with installation, at least on Solaris for me, but I can't brush off DB2 like I did Informix. If DB2 has a downside, it might possibly be that there is too much documentation. There are thousands and thousands of pages of docs and just finding the manuals can be tough (Find the DB2 v8 manuals here). One thing I noticed is that lots of docs and tutorials refer to a sqllib directory in the instance user directory... and I didn't have one. Odd, I thought, but kept on looking since maybe that was refering to an older version of DB2.

   Then I found this post at DBforums which showed how to create all the proper files including sqllib for an instance user. Ah ha! That sounded right, so I tried it ($db2_install_prefix/instance/db2icrt -u db2fenc1 db2inst1) and presto chango!!! Had to run that command as root, so I then su'ed to db2inst1 and looked at the output of env, finally all my enviromental variables were set and things smelled right. Tried db2 and everything worked. Now I can start and stop DB2, I can run the utilities, everything is fixed.

   So in the end it looks like my real problem was with the setup of the instance user (db2inst1). The db2setup installer must have missed this or bombed quietly on it and I've been chasing my tail since.

So the good news is that I've got the latest release of DB2 up and running and I'm not playing with the database. So far, I'm really impressed but we'll see how things progress over the next couple of days as I get more and more familiar with it. I'm really happy with IBM's trial support, top notch. They didn't solve my problem but they really went the extra mile and I'm convinced that given enough time of back-and-forth they'd have worked it out. So the only real problem is that the installers made some b00-b00's that required me to start fixing my installation before I'd even gotten to start using it. This is, at the least, poor form. I'm not sure if other Solaris users are having similar problems, but I've done this install now on 2 diffrent systems with the same results. Oh well, its all fixed and I'm happy now. I'll post a "Maybe I don't hate DB2 so much anymore..." blog entry in a couple of days when I've had some time to tinker.

Update 2 (48 Hours Later):

Here's a "slap yourself around" moment. I was curious why db2setup didn't setup the db2inst1 account properly but once I ran db2icrt manually, like I mentioned above, I didn't give it much more thought. But my main man Mr. Rudy at IBM mentioned when closing the support case that I ought to have a look in the installer log (/tmp/db2setup.log) just to find out why it really failed. d0h. Didn't think of that. Stupid Stupid Stupid. When I look in that log and search for "icrt" I find this:

Command to be run: "/opt/IBM/db2/V8.1/instance/db2icrt -a SERVER -s ese -u db2fenc1 -w 64 -p db2c_db2inst1 db2inst1".
ERROR:An error occurred while creating the instance "db2inst1". The return code
is "1". Create the instance manually using the command "db2icrt".

ERROR:DBI1030E Kernel parameter MSGMAX must be set to
          65535.

Explanation:
DB2 requires certain kernel parameters to be updated.

User Response:
o   Update all necessary kernel parameters as documented in the
    "Quick Beginnings" manual.
o   Reboot the system
o   Try this command again

d0h! I never saw the kernel parameters in the 'Quick Beginnings" manual. I actually needed to change two parameters: MSGMAX and MSGMNB. The funny thing is that while iCRT failed during startup, I got DB2 up and running manually without having set those parameters. So I took a look in the 'Solaris Tunable Parameters Reference Manual' and apparently both tunables are obsolete in S10. So I can only assume something being called by the db2icrt script is looking for the kernel parameters itself, not finding them, and stopping before it even tries. Without scrapping my existing instance I can't re-run the command to check it out, but perhaps next time I do a DB2 install I'll take a closer look at it

Anyway, thats the scoop. Based on what I'm seeing here, I shouldn't be the only guy in the world having problems with DB2, this should be a widespread DB2 issue on Solaris 10 and OpenSolaris. Hopefully future users that run into these brick walls will find this post and benefit.

As a side note, its all too possible that the problems with the CLI DB2 installer (db2install) are related to this kernel parameter issue.

People have asked me why I have this loathing sort of hatred of IBM before and I've always had trouble answering it because the answer is somewhat self evident. When I use an IBM product its like I'm magically and mystically swept away and taken to a magic fairy land known as the early 1980's. Mainframes buzz, magnetic media meant big ass tapes, and ADMINISTRATIVELOGON seemed like a sensable username. Ah yes, the wonderful world of IBM. Maybe no one bothered to tell them that this is the 21st century, but so be it. But I'm supposed to be talking about Informix...

So, I downloaded Informix Dynamic Server 10.0 and installed it on my beloved Sun Blade 1000 (Thank You Anysystem!!) running Solaris 10 Update 1. The installer is a pile of crap, but the install happens easily enough if you read and carefully follow the prompts.

Once Informix is installed, its time to set the baby up. Except, its not all that straight forward, you've gotta dig around the docs to find it, when you do, you realize that you've got to hand-edit two files: sqlhost and onconfig. So you copy those over from the default .std files and start editing. After hitting Google for some checks and the format of the sqlhost file I've got it hacked up and ready to go. So... now how do I start this thing? More digging, find out you need to use oninit -i to initialize the thing. Oops, doesn't work. Apparently you've got to manually allocate the datafile within which the base database information lives (in Oracle we'd call this the catalog datafiles). Okey, so thats allocated.. a big fat $50,000 database and I've used 'dd' before I even got the thing up and running. Try oninit -i again and it works.

So now apparently Informix is running. What now? More manual digging. I see something about this apps called dbaccess, but after reading a little and then trying it I determine that its another of IBM's "Wyse terminals are due for a comeback!"-isms. So I look and look and look for some other way to get an interactive SQL interface to the database, but no go. Apparently I'm stuck using dbaccess, oh goodie. So how do I start, stop, and manage this thing anyway? All thats done via this pile of on commands, in this case oninit to startup and onmode to change the operating mode of the database server, including shutdown, single-user, and quiescent. More looking around and I find the pile of nsr commands for Legato Networker(??), the archive tools, tools for managing dbspaces (the Informix equivilent to a datafile), and others, but still no alternative to dbaccess, although the BladeManager interface is actually pretty decent, a sure sign that it wasn't written by IBM people.

I've always loved CURSES, but IBM just has to make it their own and make it nasty and painful. After hacking around in dbaccess trying to create databases, interact with them, and other fun stuff I gave up in frustration.

So why do I hate Informix? Its a backward, "I Love the 80's", pile of crap thats in massive need of bulk overhaul. Why not create a simple and flexable interactive SQL interface like every other database on planet earth? And why is a $50,000 database unable to be shipped with a management GUI in case I want it? And why did I have to manually allocate the root dbspace? No one at IBM knows how to write a script apparently.

I'm sorry, I know someone is gonna hammer me on this and tell me that I just don't understand.. and your right, I don't. But I don't know if I want to. While Oracle 10g and Solaris 10 are cutting new ground Informix is lost in the past. I know its supposed to have some kick ass features but I can't get far enough into the product to even care. Apparently you can use tools like Informix Server Administrator (ISA) and Informix SQL to lessen the pain, but those aren't included (or avalible that I can see) for eval and their extra cost.

I just finished downloading DB2 UDB, I'll see if its less backward, but if memory serves its full of its own IBM-isms.

Thats right, me, the last guy on planet earth that you'd expect to run Windows Vista (aka: Longhorn) gave it a spin over the weekend. It was install for a less than 4 hours, 3 hours of which I was busy playing with Nova, after which time I uninstalled it.

I fired it up on my mighty home system: Athlon64 X2 4200+ (Dual Core 2.2Ghz) with 2GB of PC3200 on a MSI K8N SLI Platnum (nForce4) with a Western Digital 40GB Raptor (10K RPM SATA) serving up as the boot disk. Not a slow system by any means. I figured that if Vista was slow on my box it was going to be slow on everything else within reason.

So, first things first: The speed issue. It isn't nearly so bad as I expected. It takes an eternity to boot but once its up it runs pretty smoothly. I tried to install Battlefield 2 to give it a good graphics test, but it complained during install and I didn't care enough to fight with it. But, given the speed of Vista, I wouldn't want to run it on something less powerful than my box, its going to really force a lot of people into the upgrade curve.

There is a new boot loader apparently. I didn't tinker with it too much, but it really didn't want to play nicely with GRUB. The boot loader is, at least, pretty decent for what its ment to do, apparently allowing some fixage to be done low level without booting into Vista proper. I took a look around Google for some pointers on the setup but didn't find any that were helpful, so if your gonna try Vista, prepare to spend some time to get your boot loader fixed up if you want to multi-boot.

Visually Vista is very striking, with lots of frosted translucency, a round start button, new themes and all that. Honestly, it looks pretty decent. Windows has looked like a bloated and boring pig for years, and then XP's (or 2000's? I'm not sure which) title bars proved that there isn't antyhing that can be done in 8 pixels that isn't even more gigantic and blue in 30 pixels. The new title bars are thinned a little but are at least very visually appealing.

The new file explorer interface is a great improvement. Basically they ripped off some NeXT goodness and layered it into Explorer and it works very well. Paths are clear, jumping back and forth in the structure of the filesystem is more efficient, and they did nice work there.

Once again, there are new additions to the configuration tools and once again they've included a "Classic View" button. The new tools look easy or better or something but for someone who only uses Windows once a year or so, I get confused when things are overly simplified. If you've ever tried to setup networking on Windows and it asked you which ISP you had, and didn't present a "I'm on a LAN damnit" option, you know what I mean. The classic "Network" as a one stop shop will live on forever.

I'm glad to say that driver support has improved drastically in Vista. Everyone touts how great driver support is in Windows, but the thing is that Windows doesn't have many of them itself. On a UNIX/Linux system, by and large, you install and go. On a Windows system apparently you install, download drivers (assuming your network card is supported), reboot, download drivers, reboot, repeat. At least, nForce4 was well supported.

Along with Vista is Internet Explorer 7. As a UNIX guy I only know IE as that browser that can't view standards based pages and has a heart attack when it sees PNG with alpha channels. Supposedly IE7 is a huge improvement, and everyone using it around the office loves it, but to me its just a big Firefox wannabe, and not a good one at that. Frankly, I'm still unimpressed by IE and have no idea why anyone would use it willingly.

Some of the hidden, so called, "goodies" in Vista are really the pride of the Vista users around... but most of them are just rip offs of other windowing systems, namely OS X and LookingGlass. Kool, yes, but they aren't new or origonal at all. And honestly, most of them are the sort of whiz-bang things that are nifty but you turn off after about a week. Popular examples would be a thumbnail view of a window by mousing over the item on the bottom task bar, as well as seeing thumbnails of the windows when using Alt-Tab. You can also view all your windows at one time and flip through them in LookingGlass style. Some people will find value in these, and some will turn them off, but I think they are missing the point. The problem with Windows interface is that its too small... that is to say, most Windows applications are intended to be used full screen, and you run several of them, but this means that managing all those windows is really tough, and sure the new features improve managment of 5+ IE windows and 2+ graphic applications and 3+ editors, and all that... but why they haven't finally introduced a native virtual window manager I don't know. Add a pager, add multiple workspaces/desktops and allow the user to navigate that way rather than by flipping through a virtual pile of windows. Its the diffrence between shuffling through a deck of cards and laying them out on a table. Sure, virtual desktops would through new users for a loop, but you could turn them off if you wanted. I just don't see why this has never happened in Windows proper.

Whats important and telling to me is that at least the boys at Microsoft are actually working on things. There is change and improvement in Vista, both visable and invisable. Whether those changes are good, bad, or indiffrent, frankly, doesn't matter to me, whats important is that they are trying. Windows has appeared to stay relatively unchanged in the last 10 years and even "big" releases like XP only came off to me as a theme change and menu shuffle. Vista is the right step forward for the product. But, its also still a big pile of crap that limits the users ability to be productive and consumes far too many resources. Almost any X window manager would be more productive to use than Windows, based solely on virtual desktop support, not taking into account all the other advantages and breakthroughs that we have to provide the user with. Vista is gonna require a mass upgrade of systems to run it, which is great news for Dell I'm sure, and it'll really help Intel boost sales of the new chip line, although I'm hoping that AMD can swoop in and save these customers from the pain of Intel.

I'll be interested ot see what happens when Vista releases and how its accepted by the larger user base, by small business and grand parents and gamers. Right now, I just can't say. Some will love it, and some will hate it... lets see who wins.

Good news for any Sybase fans, Adaptive Server Enterprise version 15.0 Developer's Edition is availible at no cost for developers. They've got versions for Windows, Linux, and Solaris/SPARC. Its too bad a Solaris/X86 isn't also avalible, I'm not a big fan of Sybase, frankly, but I'd take it over DB2 or Informix any day. Transact-SQL is funky, but provides a fun diversion from time to time. If you like to tinker with databases, give it a shot.

Don't forget: Mashup Camp, February 20th - 21st, at the Computer History Museum in Mountain View, California! Some big and heavy hitters will be there, so if you've got a way with these things, make sure your there. I hear that there will be some pretty kick ass prizes too, but I can't say what they are... but trust me, definately worth it!

At this point 95% of you are scratching your head and wondering why there would be a camp for people mixing diffrent types of music together or who are really into potatos. Mashup, overused word of the year, in this case refers to the act of creating an application using two or more diffrent open interfaces off the net. For instance, using Yahoo! White Pages API and Google's Maps API together to form a single unified "mashup" app. Wikipedia defines it like this:

A mashup is a website or web application that seamlessly combines content from more than one source into an integrated experience. Content used in mashups is typically sourced from a third party via a public interface or API. Other methods of sourcing content for mashups include Web feeds (e.g. RSS or Atom) and JavaScript includes.

This isn't really my kinda coding. Most of the interfaces are Java, C++, or JavaScript, and prefer that you run Windows... none of which is for me. Oh well. I defer to my motto:

Web 2.0? Pffft. I just upgraded to 5.3 Beta.

Viva open source. If you want to learn more, visit MashupCamp.com and get all the details.

How does this look? I felt inspired today to do something bold, daring, and in blazing industrial orange (an ode to the formerly beautiful orange MP4-21). Ultimately, the "Its, ummm, like.... very very orange." feedback got the better of me. In all blaze orange it looked frickin' beautiful, but it was, perhaps, too artsie or modern or something, so I've toned it back and molded it into something doable. Let me know what ya think.

The official McLaren livery is ready and on track...

It looks nice... but personally, I'm bummed out they aren't actually rolling out with that industrial orange... it looked so damned hot. But, it still looks good. Notice that the silver on the car is actually chromish reflective paint. This is going to be a really good lookin' car on track as it glissens in the Sun.

There is one thing that I'm worried about: where is the Sun logo? For years there has been a Sun Microsystems logo on the side mirrors, but the current livery has omited it, at least on the official pictures. Whats the deal? Something we should know about? Or is this just tenative with changes to come?

Its here! Its here! Its finally mother-f'ing here!

Only a short 10 years after the book was actually needed, authors Peter Prinz and Tony Crawford have stepped up to the plate and offered us C in a Nutshell, yet another in the line of Nutshell books from O'Reilly. Frankly, this should have been the section nutshell book behind the venerable UNIX in a Nutshell, but whatcha gonna do. Really... I'm not bitter. Apparently they wanted to get 84 other (I kid you not) Nutshell books out of the way before feeling truely capable of handling this preciuos tome.

But now that its here, we can rejoice. The book covers C though C99 and focuses pretty exclusively on the standard. The advantage here is that you can use this book reguardless of which compiler you use (I know your using Sun Studio though... right?). The breakdown of the book is as follows:

1. Language Basics
2. Types
3. Literals
4. Type Conversions
5. Expressions and Operators
6. Statements
7. Functions
8. Arrays
9. Pointers
10. Structures, Unions, and Bit-Fields
11. Declarations
12. Dynamic Memory Management
13. Input and Output
14. Preprocessing Directives
15. Standard Library
16. The Standard Headers
17. Functions at a Glance
18. Standard Library Functions
19. Compiling with GCC
20. Using make to Build C Programs
21. Debugging C Programs with GDB

Presented here are three sections: Language, Standard Library, and Basic Tools. Like I said, this book is standards based, so the standard library reference, which covers both functions and headers, is based on the spec, not those wonderful GNU LibC-isms. In a addition, the basic tools coverage of GCC, Make, and GDB are excellent. I can understand why the coverage choice for a compiler is GCC, it is the most widely accessable and used compiler, however it would have been nice for me if they'd choosen to cover DBX as the compiler and then broke it down where GDB and DBX diverge.

The language coverage is really good. Its accessable to both new and experienced programmers, but naturally to hit that balance, plus keep the size of the book down, coverage won't satisfy everyone all of the time. I thought that it could have been better rounded and more general examples could have been used. The layout of this nutshell book is greatly diffrent than most because it is written more like a book than a reference, but it covers both bases.

Anyone writing in C would benifit themselves to keep a copy of this beside their favorite systems reference (I'd personally recommend Solaris Systems Programming by Rich Teer). No programmer with more than a years experience with C needs this book, but you might fill in some gaps in your knowledge by keeping it around. If nothing else, its a great book to keep in the bathroom for those great 5 minutes... um... cram sessions.

Its hard to believe that database giant Oracle is quietly gobbling up the databases that sit behind the most important applications around... but they are. The major news sites apparently haven't seen it yet, although our very own Mark Mayo is on the job as usual. The Oracle announcement is here. The official press release is here. You can even read about it in the company blog.

Still not sure what Sleepycat is? We're talking about BerkeleyDB, the embedded database that drives most everything... from LDAP to NIS to E's old config system to embedded devices, on and on and on. In almost any large stack of software you'll find BerkeleyDB in there somewhere and almost every UNIX system in the world has it installed and in use. I really hope that Oracle can be trusted with it. Perhaps this is how MySQL people felt when Oracle bought InnoDB.

This does somewhat elevate SQLite, and as a SQLite fan thats a good thing, I guess... but its a shame it came at the expense of BDB. Gobble Gobble. :(

My favorite stock market is old... but we still love her.

Get it now! OpenSolaris on Xen, the first source release is ready to fly! So tonight, as your holding your lover close to your body, whispering sweet nothings in her ear, just make sure that you don't subconciously compare her to a X4200 running 8 domains. Although I'm sure she's be flattered if you told her she was as smart and forward looking as both Todd Clayton and Tim Marsland combined... I mean, what woman wouldn't be breathless with flattery when held up against that standard?

Seriously, the crew at Sun has been working hard on this for a long time (about a year, that I know of) and has some of their best minds on it. Soon we'll have virtualisation bother above and below a running kernel for a truely end-to-end solution previously only seen on big expensive midrange systems and StarFire. Congrats to the whole team for doing an amazing job! We look forward to what you'll do in the future with this project.

A big round of appause for these guys: Todd Clayton, Tim Marsland, John Levon, Mark Johnson, Stu Maybee, Joe Bonasera, Ryan Scott, Dave Edmondson, John Danielson, and everyone else involved with the project. Hopefully soon I can find a fun nitch to fill and add my name to the list. Community development rules!

There has been a little concern recently about the fact that Solaris Express: Community Release Build 33's release has been held up. It would seem that some of us haven't done an adiquate job of keeping everyone informed. The cause for the concern is in the fact that the OpenSolaris B33 code was released on schedual but due to a flag day (caused by changes in Net-SNMP which are depended upon in other parts of the OS, very possitive changes I might add!) you can't build the source unless you have build 32 or higher (see the B33 release notes for details).

This wouldn't normally be a problem, except that the Solaris Express: Community Release to coincide with the OpenSolaris B33 source release got held up. Why was it held? A very good thing is happening to SX:CR, Real Player for both X86 and SPARC is being integrated! This is excellent news for everyone. Especially when you consider that Real Player now plays much more than just Real Video, but can handle a variety of plugin codec backends, such as DiVX, XviD and others, making Real Player very very interesting to all of us! But, while the source for Real is open (learn more and get involved at the Helix Community) there are trademarks and legal things like that involved in the process. These problems are non-technical but beaurocratic red-tape, and sadly, its just something you have to deal with. Normally OS's don't have to deal with this because they just hold the release and no one is the wiser and they've got months to work it out... Solaris Express releases are flying out the door quick and its got to make life more difficult at Sun for the guys prep'ing the releases.

Personally, this isn't a big deal. This is the OpenSolaris project and the code is being released regularly and efficiently. There is much improvement to be made, for sure, but thats something we're working on. For information on that Community Advisory Board page on OpenSolaris.org and become a part of the work being done to construct our governance model. This is an evolving process and we're all, inside and out, learning how to do this better and more efficiently. Its a real bummer that two things had to come together to cause the hold up, but we've still got Solaris Express: Community Release 30 to run, OpenSolaris Build 33 to hack on, and kick ass compilers and tools to play with. The fact that you'll have to wait a couple of days to build your newest code isn't that big a deal... it happens.

Word currently is that SX:CR Build 33 should be avalible today, but please do be aware that Kernel Engineering is not in charge of these Solaris Express releases! Its an entirely diffrent group. When the releases are late the community and the kernel engineering group are both asking why together. So, there is some FUD spreading around here and there, but don't let it get you down. This is an evolutionary proccess as we learn together and grow as a community. I'm really happy with the way things are going and the way that Sun is handling changes... they really care and are doing their best, the rest is for the CAB and the governance model to work out.

As promised, OpenSolaris: Community Release Build 33 is ready for download. Go get it. I've got my stack of CD-R ready and waiting for the downloads to complete. Both CD and DVD ISO's are available.

After seeing Eric Boutilier's post about Ethereal I thought I'd hit their site and catch up on things. I use Ethereal all the time but I rarely go back to the main site. Apparently they've incorperated, which is kool. They are now offering, as a business service, formal education on the tool and starting to sell goodies. You've gotta check out the Ethereal tshirts they have for sale. Classic!

They've got 4 diffrent shirts for sale and they all kick ass. I think its time to refresh my tshirt collection. :)

News just broke about Skype and Intel making a deal to limit 10-way concalls to Intel only architectures. This is madness and utter nonsense! Its gotta stop here, and now. Its one thing to impose technical requirements on a peice of software, such as requiring an nVidia graphics card to properly run a game... but its quite another to just decide that for a non-technical reason, we won't allow certain people to use the software. If this isn't market domination I don't know what is. Intel is so damned scared of AMD that its going to do anything in its power to lock them out. While the rest of the industry becomes more open, Intel just backward.

FUCK INTEL. And shame on Skype/eBay for being so pathetic as to profit in this way.

Skype users, please consider using an alternative.

Back in November CA sold Ingres to Garnett & Helfrich Capital. In the last week we've seen a lot of action around Ingres. The newly formed Ingres Corperation is gobbling up people, get an idea on their jobs page. They aren't far from me... should I go interview? :)

Ingres is now dual licensed, GPL and a commercial license for ISV's. The GPL move has met with resistance from those embedding the database and a "Shit, we're thinking about it..." responce was issued yesterday.

Perhaps most disturbing is that all the prior releases of Ingres have been removed from CA's site. Last week I downloaded Ingres 3.0.2 for Solaris10/X86... this weekend I wanted to install it on my system at home and it was gone. All the non-Linux releases are gone, replaced with a clearly Linux-centric focus. The "new" release is dubbed "Ingres 2006 GA" (v9.0.4) and there are 3 DB downloads on their download portal page:

• INGRES 2006 GA Release for Linux—x86 32-bit
• INGRES 2006 GA Release for Linux—x86 32-bit—Certified Binaries Trial License
• INGRES 2006 GA Release for Linux—x86 32-bit—Source Code under GNU General Public License version 2.

Even the source is being labeled for Linux. Are they snubbing other platforms such as Solaris?

Clearly this newly formed Ingres Corp wants to go toe-to-toe with MySQL. Good. Frickin'. Luck. I don't think they've got a leg to stand on. They are going to have to grow up and realize that Linux isn't the only place to gain traction. Maybe this is premature, maybe the other UNIX platforms will be addressed in the future, but it sure doesn't look like it right now. Linux has become a battleground for the database, much as it was for HPC in the last several years. Oracle's aquisition of InnoDB seems to have really taken the gloves off the players. But Linux is the last platform that I'd want to deploy my database on. Solaris is built and bred to serve databases... I just can't see the wisdom here. Open Source Database doesn't have to mean Linux... Solaris has a better way, a more stable way, and a proven track record. Plus, its cheaper than Linux, has a vastly better support organization at nominal cost thats, again, cheaper than RHEL, and Sun goes further with a complete portfolio of servers and other software to fill in the solution.

But hey... who cares right. Sun knows databases! But oh well, ya, lets buy a bunch of frickin' Dell's and deploy a database. It might seem like a quick and cheap option on papers, but its just not the reality. Sun's got the value, the experience, and the solutions for databases. If Ingres wants a comprehensive solution they've got to deploy on more than just Linux.

Keep your eyes on Ingres and lets see what happens.

Al Hopper and I have kicked around the idea of putting a Wiki on Genunix for some time now, but on Wednesday's CAB call it was requested officially to be used for future editing and work on the governance documents and charter. That night I did a test setup at home, and last night I brought the production wiki online. This wiki will act as the official community wiki of the project.

Given that the wiki is about a day old, its sorta empty. I've filled in some content, just to get the ball rolling, but I'd love to see people dig there hands in and get good and dirty. Don't be shy, do what you can to help out. Its version controlled, so if you break something we can roll it back or tweek it... thats what its all about.

Visit the OpenSolaris WIki now at www.genunix.org/wiki. Please share feedback and ideas on how we can make it even better.

Enlightenments DR16 (e16) window manager continues to live and thrive in the capable hands of Mr. Kim Woelders who triumphantly took ownership of the codebase following DR16.5. The last full release of DR16 was 16.7.2 in December of 2004, so this release incorperates lots of new functionality, improvements, and bug fixes. Those who haven't used DR16 in a long time will be excited to check out the Xorg Composite (translucency) support built into the WM.

Also, Kim and I made a late breaking change to the project that will be of interest to users who are running on Solaris or other non-GNU operating systems: we've removed gettextize (and the script we used to wrap it) from the autof00 bootstrap (autogen.sh) and moved to using autopoint. This is signiicant because gettext autof00 was really nasty on Solaris requiring you to either do a lot of fixing or simply building from an existing dist (pre-generated tarball).

Congrats to Kim on a fine release! I recommend anyone wanting a kickin' stable and powerful WM check out DR16. Even though DR17 is running full steam ahead there is still a distinct place in the world for DR16 and I run it every day... I couldn't be productive without it, frankly.

Download Enlightenment 0.16.8 Here. Read Kim's release announcement in the e-devel archives.

Today was very interesting... I got to break out of the norm and be Mr. Big Shot. At least, for me it was pretty nifty. I stacked up several Sun related events and took a half-day of vacation to hang out at Sun's Menlo Park campus.

The day started off in Menlo Park Building 14 (MPK14) where I was asked to be interviewed for a segment of the upcoming Sun Developer Network Channel show, supposedly to be released around the 22nd of Feb. This was an all together interesting experience for me for several reasons. First, I'm not charasmatic before about 1pm but they scheduled the interview for 11am, so my answers were short and lackluster and I probly looked alittle spacy. In a sense that worked to my advantage because the interview was so short and he switch topics so quickly I couldn't really give a meaningful answer anyway. Secondly, because it was like all "lights-camera-action", just like you'd joke about. There was a crew of 3, camera guy, sound guy, and (I kid you not) a director, in addition to the host and co-host. In an effort to reach "real developers" they don't want these videos to seem corp-ish, so how do the solve the problem? Film it standing next to a foosball table. Ironically, I hate foosball, its just retarded and lame. "Real" would have been outside JavaJava (a little upscale cafe inside the Menlo Park Campus) having coffee and a cig. But, foosball it was. Now, to make matters more interesting, they filmed several close up shots of a goal being scored, and then they cut to the host shaking my hand and saying "Haha, great game.... My guest is Ben Rockwood, good to have you on the shot" etc. So, it might look like I'm playing a match, but I literally never touched the table. The third interesting bit, is that not only am I standing next to a foosball table and mic'ed for sound and only half awake but in order to stay in frame, because the camera was so close (filmed in a hallway) I'm like no more than a foot away from the host, and so its kinda akward, for me anyway. Very very interesting experience. Apparently it went well because we "got it on the first take". 11am I was there, 11:30 I was escorted out, just that quick. I'm totally honored that they'd ask me to do it. I think they said that my segment would come right after a segment by Jonathan, so thats even more kick ass. Maybe it'll look good, I don't know. I'm a minimalist sort of guy that figures "real" means coffee and smoking or Guinness and laptops... you know, like "real" geek stuff. But maybe those ideas are to be saved for my own show, should I ever make one. :)

After the filming (sounds wierd to say) I had an hour before my next stop so I asked Sun's best and brightest PM, Jennifer Bauer, if she'd be free for an early lunch, and thankfully she was. We had a nice lunch in the caf and caught up on things. I always enjoy keeping up on her action packed adventures.

Following lunch it was off to MPK17 to join in on the second half of the CAB Con-Call. It was fun though to be on the call in person with all the Sun guys. I looked for some of the guys that would be on the call in their offices but they weren't around, so I checked a couple confrence rooms and eventually, not finding them, just ducked into an empty room and called in. Turns out they were down a couple halls for me, so I gathered my stuff and ran over. To my supprise both Simon and Casper were in the area and in the room for the call, along with Jim and Stephen, leaving Rich and Al on the phone.

At 1pm I was done with the call and was tenatively schedualed to meet with Bill Vass (Sun CIO), or one of his people, about an article for an upcoming newsletter. How kool is that! Sadly, I'd tried all morning to contact the appropriate persons and just couldn't get ahold of them, so after leaving another vmail or two I figured it wasn't happening today. So I guess that'll be forthcoming.

I figured since I was on campus I'd try to slip into the Sun Store on campus and pick up some new Sun shirts and stuff. I walked over, rockin' out to some Fear Factory on the iPod, and into the caf when it just so happened, by chance, that I'd be walking the whole time behind two very kickass persons: Dr. Stephen Hahn and Karyn Ritter. I'd never met Karyn in person before, and the first time I'd met Stephen in person was in the call that I'd just been on, although I didn't realize it was him from the back. When they saw me behind them they said hello and invited me to have lunch with them. Thats an invite you just don't say no to, but I'd eaten so I told 'em that I'd catch up to them a couple minutes later. So I ran into the Sun Store, bought about $110 worth of Sun shirts, then ran over to get some coffee from the Pete's they have there (although no one was behind the counter, so after a minute or two I freaked out the other people around by just going behind the counter, pouring my own coffee and leaving $2 on the register).

I had a wonderful chat with both Stephen and Karyn. Both of them are exceptionally bright people and very friendly. Stephens personality is pronounced, very bold, and very powerful. Karyn was very kind, very experienced, and has a classic smile thats friendly and warm, a trait which reminded me very much of Claire Giordano. Our discussion was interesting and lively. I'm always interested in getting off-the-record, personal, and open input on how they feel the community is doing, what we can do to help, and just get general feedback and opinions that you just can't get via a more structured communications channel like email. Lots of ideas, comments, and experiences flew back and forth and it was thoroughly enjoyable and enlightening. Its one of those sorts of discussions that you don't neccisarily walk away with any specific point, per se, but you tap the pulse of the project from specific points of view, high ranking internal ones in this case, and get a general vibe on things. I consider the conversation private, and thus won't disclose the contents of the discussion but I'll say that one thing I look for in discussions like this is whether or not the vibe I get from them in person is consistant with the vibe you get from them in general... that is to say, are they geniune or do they start unloading on people when the cameras are turned off. I'm glad to say that without a doubt, both Karyn and Stephen are straight shooters, very consistant, and very driven, and I feel even more confident in the project as a whole with them holding the reigns internally. It was a great oppertunity and I'm really greatful for it.

We wrapped up lunch about 2:20pm and I ran back to my car. Since I actually have a job that isn't with Sun I didn't want to get into hot water with my manager, particularly because I'm in the middle of several time sensative projects right now. Thankfully he was understanding about it all and there wasn't a problem. The rest of the afternoon consisted of email, finally getting a media kit for NetBackup 6.0 and installing it on my bad ass Sun X4200 server, and prep'ing 6 crappy X86 boxes for a mail system I'm eval'ing.

So, a very interesting day. Being around so many important folks and being featured in a video and all that kinda makes me look like a big shot guy, but I'll tell you I don't feel like it. Its both intimidating and exciting to be involved with all these things and to have your opinion valued and requested. All these people at Sun are just people, like you and I, but as a Solaris enthusiast and Sun zealot I look up to these people with great respect, to hang out in the same room with them or having lunch with them is no small thing in my book. I certainly don't take anything for granted. I feel really honored and greatful to be involved as much as I am and that I'm given so many oppertunities to be involved. I'm no one special or particularly unique, no more so than anyone reading this blog, its perhaps due more to the fact that I'm a) very vocal, and b) live in the silicon valley where all this stuff happens. If you (my humble reader) and I swapped places I'd be the one reading this post, and so I might get to play Mr. Big Shot for a day, but its just a day, and its not solely because I'm me, but, it sure is fun for a day.

Why... oh why... my beloved 3800, one that was to be mine, and yet... wasn't. Why!?!?! Oh why?

Behold, the Sun Fire 3800 Enterprise Server. The system went GA (general avaliblity) on April 16, 2001 and, only a short time later, in August 2003, was EOL (end of life) with the final shipment taking place in November of that same year. Only two years and then *p00f*. Gone. By contrast, the system that was so bold and magificent as to capture my eternal praise, the Sun Enterprise 4500, GA'ed in April 1998 and EOL'ed January 2003. E3800 was introduced as part of the Serengeti line of servers, the bold new 3rd generation of 64bit midrange systems that was the final transfer of Cray's engineering in the Starfire Enterprise 10K server, down into the midrage systems. The Serengeti range of systems debuted with E3800, E4800, E4810 (which was a rackmount version of the freestanding 4800) , and E6800, in addition to the V1280. In 2003 the x800 servers were phased out and EOL'ed to be replaced by the x900 servers, and uplift of the existing line, with one little switcheroo, the 3800 was removed from the lineup and repalced with the 2900, which was an uplifted V1280. The big unanswered question for me is, why?

So why love the 3800 you ask? Serengeti really put the ball back in Sun's corner, catch up to and surpassing what IBM was offering, at a much lower price point. The systems now all has FRU service processors (SP's) which was effectively a very sophisticated LOM from the Netra side of the world. You could telnet into the service processor of a system that was powered down, check the event log, run diagnostics on the hardware, check system status, all with the system powered off, then you could keep careful watch on the system when it was up, all from simple telnet session, no console servers required. The backplanes, now dubbed Fireplane, was sophisticated and blazingly (no pun) fast. The CPU/Memory boards were really solid and well built, not like the flimsy and fragile Ex000/Ex500 boards that just squeezed into a thin metal slot. The system was completely redundant, including redundant repeater boards, truely hot-swap just about everything (try hotswapping on an E4500... please... you can swap it in, but not out... at least, the swapout is otherwise refered to as a "reboot"). And the feature, o features.... Dynamic System Domains on the SP! w00t. You could carve up your system into multiple independant systems, something previously only avaible on the Starfire systems. Ya, Serengeti was (and still is!) the ultimate realization of Sun's legendary midrange systems lineup.

But I still haven't explained why the E3800 was diffrent! Well... the E3800 has one thing that its big brothers didn't: cPCI. And whats so great about compact PCI? Its hot swappable. But isn't IO hot swappable on other Sun Fire sytems? Yes... and no. On other other Sun Fire systems you can't hot swap a PCI card, per se, but you can hot swap a "PCI Cage". Each PCI cage would have a number of (usual 8, exept for V1280/E2900, with has only 5) PCI slots. You'd swap out the whole cage as a single FRU (field replacable unit). Typically this is a bad thing. Now, if you architect for total redundancy you can use 2 identically equippted PCI cages with, say, 2 4Gb FC Fabric HBA's, 2 GigEthers, and maybe Diff SCSI, and then in software make all these redunant (IPMP, MultiPath Fibre, etc) and the hot swapping the whole cage isn't a problem. Unless, of course, you have V1280/E2900, in which case you've only got a single PCI cage. Anyway, cPCI allows you to swap just a single card, and thats a good thing. Sure, we can debate technically the problems inherent to cPCI, including the limited availbity of decent cPCI cards, but damnit, it was the way to go. At the time that E3800 was avalible you could also order cPCI IO bays (cages) for the other servers in the line but apparently when the E3800 went, so did cPCI as a whole, suggesting that it was cPCI that brought about the demise of the system.

The E3800 was also just a cute little guy. Front accessable, well layed out. Gotta love it. Sure, it doesn't have redundant system controllers (SP) but then the V1280/E2900 doesn't either. Damn. I guess it just wasn't to be. But damnit, I'd sure as hell still love to have one.

Now... in Sun's defense, swinging the V1280 into the line with E2900 was a good move, although I don't think it should have been at the expense of E3800. I was one of Sun's first customers to get a V1280, a pair actually. We ordered ours several months ahead of GA, and I was nervous about buying a system I'd never seen and had no experience on.... but wow, what a system. The design of the V1280 is... its... you can't put it into words. Its unique, its bold, its distinctive, and its just damned solid engineering! When I first unpacked the systems I couldn't help but think of a Boeing 737, not an inch of space is wasted and its exceptionally efficient. The V1280's boards are top loaded! This allows maximum efficient of the airflow from front to back, rather than most midrange systems which push air from side-to-side or top-to-bottom. This allows you to cram a lot of these babies into a small space, one on top of another, and still have fully adiquate cooling capabilities. You'd think that putting boards in would be scary because they'd slam into place by the force of gravity, but here again good engineering saves the day with "anti-gravity" slides, so that you can slide a board half way into the slot, let go, and it just sits there, it doesn't fall. To keep the handles efficient, they are spring loaded, insert a phillips screw-driver, twist and the handles pop up. Good damned engineering. But... here is the problem. If you rack mount this thing, you've got to slide the entire system out of the rack to pull a board. This is one heavy mother-f'er.... you've gotta really have secure racks to hold this thing if you expect to slide it out. Sadly, in my case, my V1280's are sitting on shelves in Chatsworth 2 post racks with simply nothing above them, which is a waste of space. Even still, the V1280/E2900 is simply a marvel of engineering and I think is perhaps the most exciting of the current Sun SPARC line of systems.

So, the question is... what as so bad about cPCI, the thing that really made the little server that would be king, that it made a really great server just fade into the history books with nothing but a muffled whimper? Someone's gotta know. I'd love comments from anyone with the inside scoop.

Lesson of the week: Kids and Keybards don't mix... particularly when that keyboard has focus on a terminal logged in as root.

Ya, thats right, I did something stupid. My son, Glenn, couldn't sleep the other night and I was getting Novell's eDirectory setup. I couldn't start the directory server due to an unfound library dependancy, and he was screaming in the crib (couldn't sleep) so I grabbed him and plopped him down on my lap while I fixed all the depenancies. Using a terminal logged in as root, I used ldd to find the list of shared libs it was looking for, found the missing lib, used find to find the lib and resolve the dependancy. Things were looking good... and the, with a flash, I hilited a string and pasted it (classic X middle-button paste)... except, apparently I moved my cursor last second and instead of hiliting just the filename I wanted I got the entire ldd output and it all got pasted line by line. This has happened to me before, and since these aren't recognized commands you just get a slew of "command not found" lines, except this time one of those lines did something. My Sun Blade 1000 started screaming, and suddenly a white area appear on the screen. Shit... OBP. I sync'ed the box from OBP and then tried to boot back into the OS, but wouldn't you know it, init/svc somehow got massively f**ked because I couldn't even boot single user mode. Result? I booted an OS install disk, verified that at least the filesystem wasn't corrupt, check things, didn't see anything out of place, and after determining that I wasn't going to find the cause in any short order I tried an upgrade to overwrite the installed OS. Glenn was having a good enough time though. Ultimately I couldn't upgrade because the installer wanted a backup medium which I didn't have, so I installed fresh to a second disk in the box, then started rsycn'ing user data into the new OS. Fun stuff. Lots of fun. Result? My mail was down for 2 days and I've been exchaning quite a bit of mail with folks.

Glenn's apparently got his mothers mystical ability to crash boxes with his mind, or mere presence. Nova and I had a similar experience months ago while I was working on a NetApp Filer. She grabbed for the keyboard while I was editing /etc/rc and then the box rebooted... result, unbootable Filer. That sucked too. Ended up having to go on site and fix it on the console, making a 5 minute change a 2 hour ordeal.

So, kids.... their cute, cuddly, and wonderful... but make sure you lock the damned console before you put them on your lap. :)

There isn't much point to this post except maybe for any of you that have recently done something really stupid, well, now you know that I'm right there in "What-were-you-smoking?!?!"-ville with you. :) To make yourself feel better I suggest Strong Bad Email and Arj and Poopy cartoons.

UPDATE: My friend "noyb" from #opensolaris (IRC) is the happy winner, correctly identifying the problem:

benr@anysystem lib$ find . -type f -size 0 -print
./libm.so.1
./libc.so.1
./libresolv.so.2
./librt.so.1
./libw.so.1

d0h!

On the wise advice of Craig Morgan I've setup two Sun Ray 1's in a multihead group. Craig is definately my kinda guy, I had no idea that Sun Ray's could do this. Its easy too, you enable the feature on the server, then login to the first Sun Ray, startup the multihead utility, create a group and define the size of it, then you log into the second Sun Ray and it adds it to the group, after which you just save it, log out and log back in. Presto, dual head. To enable Xinerama you use the utxconfig utility. I went a litlle further and hardcoded my resolution since the auto setting was bringing my Sony 17" CRTs as 1024x768.

Here you can see 2 Sony 17" CRTs connected to 2 Sun Ray 1's. You can see on of the Sun Rays, which is wired up with power, ethernet, video, keyboard, and mouse. The other Sun Ray is sitting behind the two monitors, and has only power, video, and ethernet. The resolution on each head is pumped up to 1280x1024, and you can Enlightenment DR17 running.

The immediate question is: how slow is it? Believe it or not, its pretty damned fast. Things that refresh really quickly suffer, such as video or listing large directories in a large terminal, but not nearly as slow as you'd expect. You really take a hit when you watch a video that is being rendered on both heads, but it works well enough. I'm sure if I dropped back down to something smaller like 1024x768 it would be significantly faster. But E is fast and responsive and the user experience is unimpared, like I said, its only very large very fast refreshes that make you realize your not on the console.

I'm impressed with these Sun Ray's. Really really impressed. It makes me happy. I take some pride in the Sun Ray's because when I was working for Sun in Sunnyvale I remember when I got requested to setup several systems for a project called "Corona". I had no idea what it was, and it was apparently very hush hush, even around the office. A couple days later these boxes are piled in front of one of the senior engineers door, about 20 of them. Then, the next day there are like 30 more. And more, and more, and more just kept getting delievered. All of the systems guys were naturally dieing for a peak in those boxes. What were they? Project Corona was dubbed Sun Ray, great successor to the not so great JavaStation, which was interesting because months prior when I worked in Menlo Park I was tasted with clearing the JavaStations out of the 3rd floor of MPK17, most of the users nearly threw the thing at me, and only one guy was pissed that I took it, and only then because he was using it as a door stop (literally). I didn't have any involvement with Corona except setting up a bunch of E4500's for what I asume was testing, but never the less I take pride in the fact that I got to have even the smallest of dealings with such an amazing project.

Also, I realized after I posted yesterday that I didn't mention what server I'm using... I'm searching the Ray's using Sun Ray Server 3.1 running on my Athlon64 X2 2.2Ghz (dual-core) box (built from parts, go NewEgg!) with 2GB's of memory, 10K RPM SATA root disk, and a bunch of SATA/PATA additional storage. The nice thing about this box is that it can do anything I could want it to and if runs slow on this box I know it won't run faster anywhere else. The server OS is, of course, OpenSolaris Build 27.

My good friend and E developer Nathan "RbdPngn' Ingersoll sent me a gift, 4 Sun Ray 1's in excellent condition. I've been contemplating buying a couple for some time and so I'm stoked to have 'em. I've known all about Sun Ray's and have advicated them for years but I've never deployed them myself. Now, finally, I'm happy to say I've got a Sun Ray in Cuddletech Labs (ie: my den) and I've turned Tamarah onto it.

Many people have asked me in the past about whether or not Enlightenment DR16 and DR17 work with Sun Ray, and I'm happy to say YES! There isn't any reason it shouldn't, of course, but now I've actually tested and used it for myself. Both window managers run beautifully on the Ray. Setup was a snap because I already use both window managers on the system that acts as the Sun Ray session server, so I didn't have to configure anything at all.

All in all, I love the Sun Ray. My only complaint is for the documentation. The manuals suck ass. They are confusing and non-linear, or perhaps I should say that they are linear but there are so many options or diffrent configurations that you skip back and forth in the manual and can easily skip a step or get lost, making it feel very non-linear. The short of install is simplifed roughly to the following:

1. Download and Install Sun Ray Services 3.1
2. Reboot.
3. In /opt/SUNWut/sbin run utconfig. Setup the framework and the web interface but don't bother testing or looking at the web interface, you don't need it (at the moment anyway).
4. Reboot.
5. In /opt/SUNWut/sbin run utadm -c to setup the framework for clients.
6. Run utadm again, but this time with the -A option followed by the your network, in my case "10.0.0.0". It'll give you a list of defaults and ask you to approve it. Here is the trick, say no! You'll then be asked all the questions one by one, when it asks you to supply a range for the DHCP clients, do so, and give it plenty of room. By default the DHCP will be configured but won't have an address range, which will leave scratching your head as you stare at a Sun Ray that won't do anything but with there with an hour glass and 2 bars under it.
7. Ensure that any pre-existing DHCP servers on your network have DHCP disabled! I have to turn the DHCP server off on my Linksys router and then power cycle it in order to get it properly disabled.
8. Now run utrestart.
9. At this point your Ray will either boot up or sit there doing nothing. If it sits there, you probly need to fix DHCP. On my box (OpenSolaris B27) I had to manual restart the DHCP server (svcadm restart dhcp-server) every time the system booted.

   Thats the short of it. The manuals are useful, so reading them is a good thing, but if it weren't for being able to ask Nathan a question here or there I'd have been up a creek for a long time till I figured it out.

   So, two thumbs way way up for Sun Ray. I've got one deployeed, now I'm thinking up possibilities for the other 3 I have. The next step is, of course, to find some cheap smart cards so I can play with these things as they were intended to be used, so if you've got a tip of cheap smart cards, preferably that are pretty and painted rather than just ugly white things, let me know.

Backing up large amounts of data isn't easy and it certainly isn't a speedy thing. For a long time I searched for a real solution for backing up several terabytes of data without letting a single backup run for a week (literally). It was LTO Gen-3 that answered the call. Finally, LTO-3 offers high speed and high capacity in a low cost solution, and Sun's (Quantum OEM) StorEdge C4 library rises to the data challange better than any other midsized library in the industry today. But... LTO-3 can suck up data at a blazing 576 GB/hr (Up to 80MB/s), which means I need a pretty agile and hardy system just to handle that amount of data flowing in from the network and then spit it onto tape. What system to use? I choose to go with the new Sun Fire X4200 2U server from the new Galaxy line.

My shiny and beautiful X4200 is top of the line, featuring dual Opteron 275's (Dual-Core 2.2Ghz), 4GB of memory, and all the IO you can eat. The quad onboard copper gigabits take care of the network and the PCI-E and PCI-X provide a place for my SCSI Ultra320 adapters to attach to my thirsty C4. The onboard SAS drives provide all the local disk I need and redundant field replacable parts like hot-swap fans and power supplies make life smell sweet again. But what I find more interesting, personally, is the Integrated Lights Out Management (ILOM).

Sun introduced its customers to the LOM back with the Netra line of telco-grade servers several years ago. Suddenly customers started looking long and hard at putting Netra T1's in their datacenters instead of down a man-hole. It wasn't long before the Seringeti line of servers brought together the combination of the Starfire's Service Processor (SP) and the Netra's Lights Out Management (LOM). This was a desperatly needed advance in enterprise servers. Every year I go to Los Angles for Christmas and when I'm there I would be reminded that there was only one thing I couldn't do on my servers: push the power button. LOM put the power back in my hands in the case of a system shutdown or post-power outage situation.

Naturally, when I unpacked my X4200 the first thing I wanted to play with was the ILOM, and I'm happy to say that it finally represents an X86 LOM solution that an old SPARC guy like me can be proud of. While its true that the Sun Fire V20z server had a LOM onboard, it left a lot to be desired. I remember being embarroused when Al Hopper and myself setup the V20z's that power Genunix.org. We were pressed for time so we racked the systems, powered them up, got responce from the LOM and triumphently said "Kool, we're done, we can do everything via the LOM." That night I realized that I hadn't setup the consoles completely and couldn't do it via the LOM, and had to make a really regretful plee to ICS to let us back into the site so that I could finish the job. The LOM on the V20z was better than nothing, but didn't stop dreaming of the V1280's SP.

If you've never used the midrange SP's your really missing out. The Service Processor on my V1280's are just amazingly awesome... but I'll talk about those in another entry if your interested.

Back to the ILOM on the X4200... Its not like the SP's on the midrange systems, but its got its own perks. The point of the LOM is to provide access to the system via a daugherboard via RS-232 console and 100Base-T ethernet at any point that power is trickling into the system, reguardless of whether the system is running or not, hense the whole "Lights Out" bit. When you plug in the X4200 the ILOM boots up immediately and about 40 seconds later you can contact it via a console, over the network via SSH or HTTP.

You can login, by default, with the login "root" and the password "changeme". The web interface is pretty and easy to use providing plenty of information about the system, the ability to change the state of the system (power on, soft reset, hard reset, power off, etc), and even to access a Java powered KVM-over-IP.

While the web interface is pretty and handy for you manager, the CLI is much more my style, so a quick SSH and you've got yourself a prompt:

$ ssh root@betty-lom
root@betty-lom's password:

Sun(TM) Integrated Lights Out Manager
Version 1.0
Copyright 2005 Sun Microsystems, Inc. All rights reserved.
Warning: password is set to factory default.

-> help
The help command is used to view information about commands and targets
 (snip)
help targets  displays a list of targets
help legal    displays the product legal notice

Commands are:
cd
create
delete
exit
help
load
reset
set
show
start
stop
version

The ILOM, to me, feels a lot like OBP on SPARC systems. You can navigate between componants of the system like a filesystem. Each componant of the system is refered to as a target. To view the status of the system, you can use the command show /SYS.

-> show /SYS

  /SYS
    Targets:
        FIOBD
        FT0
        FT1
        MB
        PDB
        PS0
        PS1
        SASBP

    Properties:
        ACT = on
        FAN_FAULT = off
        LOCATE = off
        POWERSTATE = on
        PSU_FAULT = off
        SERVICE = off
        TEMP_FAULT = off

    Commands:
        cd
        reset
        show
        start
        stop

You can see here that the /SYS target has several targets beneath it (and several yet below them) such as MB (Motherboard), PS0 (Power Supply 0), FT0 (Fan Tray 0), etc. You also see a list or properties for that target, in this case we can see that the POWERSTATE is on (its running), the LOCATE light is off, TEMP_FAULT is off meaning that the temp is ok, etc, etc. Finally, each target when show'ed will list the commands that can be used with that target, in this case we can start, stop, and reset the target. If you were to use one of these commands on the /SYS target it would effect the whole system, so start /SYS will powerup and boot the system and stop /SYS would power it off. Various arguments can be passed to ILOM commands, like -f to force a start or stop. By default, stopping the system will be done softly, allowing time for the OS to shutdown and halt before finally removing power, forcing the operating will still send a signal to the OS but won't wait for it to halt before removing power.

The ILOM itself is configured and manipulated via this target system.

-> show /SP

  /SP
    Targets:
        alert
        cli
        clients
        clock
        console
        logs
        network
        serial
        services
        sessions
        users

    Properties:

    Commands:
        cd
        reset
        show
        version

Here we see a variety of targets provided by the SP (ILOM), such as access to the system console (/SP/console), ILOM users (/SP/users), etc. If you wanted to check the network settings of the ILOM itself, you could do it like this:

-> cd /SP/network
/SP/network

-> show

  /SP/network
    Targets:

    Properties:
        commitpending = (Cannot show property)
        ipaddress = 10.10.0.134
        ipdiscovery = Static
        ipgateway = 10.10.1.1
        ipnetmask = 255.255.252.0
        macaddress = 00:14:4F:0E:xx:xx
        pendingipaddress = 10.10.0.134
        pendingipdiscovery = Static
        pendingipgateway = 10.10.1.1
        pendingipnetmask = 255.255.252.0

    Commands:
        cd
        set
        show

I'll admit that when I first used this ILOM I was really disappointed and bothered by the fact that its not more straight forward like the midrange SPs, but I'll admit that once you get used to it and the targets don't seem so foreign and confusing (ie: you start to know where to find things) it all becomes pretty comfortable. I'm not saying that its as good as the midrange SP's, but its by far the best X86 LOM that I've ever seen. Here is an example of checking the SP event log, then booting the system, and then connecting to the console:

-> show /SP/logs/event/list

  /SP/logs/event/list
    Targets:
    Properties:

    Commands:
        show

    Events:
        EventId                 TimeStamp       SensorName  SensorType
              1  Sat Dec 10 03:40:02 2005        ps1.pwrok  Power Supply

             10  Thu Jan  1 00:00:54 1970        ps1.pwrok  Power Supply

            100  Thu Jan  1 00:01:12 1970        ps0.pwrok  Power Supply

            101  Thu Jan  1 00:01:13 1970        ps0.prsnt  Entity Presence

            102  Thu Jan  1 00:01:14 1970        ps1.vinok  Power Supply

            103  Thu Jan 19 17:49:41 2006          Unknown  BIOS POST Progress
                ....

-> start /SYS
Are you sure you want to start /SYS (y/n)? y
Starting /SYS

-> start /SP/console
Are you sure you want to start /SP/console (y/n)? y
To stop, hit ESC (

Sun Fire X4200 Server, 2 AMD North Bridges, Rev E6
1 AMD 8111 I/O Hub, Rev C2
2 AMD 8131 PCI-X Controllers, Rev B2
System Serial Number  : 0550AM1847
BMC Firmware Revision :  1.00
Checking NVRAM..
Initializing USB Controllers .. Done.
Press F2 to run Setup (CTRL+E on Remote Keyboard)
Press F12 to boot from the network (CTRL+N on Remote Keyboard)
Press F8 for BBS POPUP  (CTRL+P on Remote Keyboard)
....

In case your wondering, the ILOM itself runs Linux. It takes a little while to boot (about 40-60 seconds) but is fast and responsive. It works well reguardless of what OS you install on the system. Origonally I ran S10 Update 1 on this box, but when I realized that Veritas (assholes) don't support Netbackup Server for Solaris/X86 I had to install CentOS, which runs like a beauty on this box.

After having owned this X4200 system for about a month now I look at it less like another server and much more like an invaluable tool. Its so fast and easy to work with that I can quickly change out OS's, setup evaluation enviroments for testing on Linux (CentOS) or Solaris, and I'm confident that even Windows would run well, or as well as a pile of shit OS can.

If your the curious type, like I am, and you want to watch the ILOM itself boot, I've put up a console log of the ILOM itself booting up. From that you can get a lot of juicy details on how its implemented.

Without doubt, the X4200 is the first X86 system on the market that I've seen that really lives up to the Sun logo on the box. Enterprise class, from end-to-end, without compromise and all with an amazing price point. The system is unbelivably loud (that very satisfyingly macho sort of loud!) and amazingly cool running. It sat, literally, next to me for 2 weeks and was never warm to the touch... just very loud. Clearly, the X4200 is my X86 server of choice and I can't wait to see the Niagra Galaxy systems that'll come soon.

A couple blocks from my house is an interesting place called Archery Only... you can kinda guess what they do. I'd stopped in a couple of times to check it out and was really impressed, they've got a full shop full of archery and hunting gear, as well as an in-door archery range right there in the shop. In fact, if you've watched MythBusters you've probly seen the range before.. they've tested several archery and crossbow related myths there. Anyway, it seemed like learning something about archery would be fun, since I don't really know more about archery than whats obvious, and it turns out that Monday nights they have a lessons night, which is thankfully two 1 hour periods. So about 3 weeks ago I went to the 7pm-8pm slot, and then Tamarah and I swapped the kids so that she could do the 8pm-9pm slot. We both had a blast and really got into it. We got to use both recurve and compound bows. Whats more, I now know what a recurve bow actually is. :)

When I got home I had to learn everything about archery that I could and became amazed by how much development there has been in the field. The bow above is the Archery Research AR-32, and its one scary mother. I've always considered myself a rifle man, personally, I enjoy marksmanship and being out on the range, but nothing quite feels more manly that holding a compound bow at full draw.

The guys at Archery Only are really kool and easy going. The guys that frequent the range are really helpful and like new folks, so its an ideal enviroment in which to learn. Tamarah and I both keep trying to find time to go back but life is busy and the place closes when the kids go to bed, making it hard to get over there. In contrast, its easier for me to go karting because the track is open till midnight, allow me to put the kids down, spend some time with Tamarah while its quiet and then slip over to the track for a couple laps.

Anyway, we here in the Silicon Valley are really lucky to have some really fun things to do. If you live here and are looking for something fun, exciting, and inexpensive to do here are some ideas:

• Archery Only: Spend some time with fun people and hone your archery skills, or just visit a MythBusters landmark.
• LeMans Indoor Karting: For only $25 you can be putting on your driving suit and helmet and buzzing around a track at 30mps only an inch off the ground. Every Sunday night they run "Sunday Night Heats", where as many as 40 drivers meet up to go head to head for the night.
• Driving the back roads: I love my Volvo S70 T5 and I love driving backroads at night, especially on one of California's warm summer nights. Niles Canyon and beyond it, Calevarous Rd which runs behind Mission Peak in Fremont, provides a wonderful mix of well maintained backroads that are almost completely empty at night, have no pedestrian traffic, are beautiful, and are rarely patrolled by police after about 10pm. The southern stretch of Calevarous Rd is highly technical and fun to drive, although tight and dangerous, featuring WRC like tarmac for those who have AWD rally grade cars.
• And like, lots more that I can't think of atm....

I thought I'd see if SHIFT had any new 2006 karting gear and, much to my supprise, couldn't find any links relating to Karting period! Some of the karting images are still around, but even in the webstore, if you click on a karting link it takes you swiftly to an MX page. A quick call to SHIFT customer service (and a callback) gave me confirmation... SHIFT is out of the karting business. A real bummer.

SHIFT, if you don't know, is a division of Fox Racing which specizes in Motorcross, Street, and untill recently, Karting. Fox, and thus SHIFT, is a master when it comes to branding, selling far more Fox logo shirts and clothes than it does actual safety gear. The SHIFT karting line was good stuff, not the top of the line in terms of quality (my driving boots have a big ass hole in the side which appeared only 5 months after having bought them), but was priced reasonably and had a very large brand presence in the sport. I suppose that leaves an opening for Alpine Stars to step into. I actually bought a SHIFT driving suit last year but returned it because the sleeves weren't long enough, even on the largest suit they made. I thought about instead buying an Alpine Stars K-MX 5 suit (its a damned nice looking suit) but instead decided to save the cash and just keep putting up with the suits provided by the track. In about a year I'll finally have to replace my boots and gloves, and I'll probly go Alpine Stars for both, although I have loved my Sparco gloves. We'll see. I don't get to the track nearly as much as I'd like to these days.

RIP SHIFT Karting. Wish I'd picked up a couple shirts before you went down in flames.