Tuesday, December 19, 2006

SLES 10 Upgrades: done! So what happens now?

This quarter we Webmasters have been focusing our attention on upgrading our servers to SLES 10 - SuSE Linux Entreprise Server 10. We were running SLES 9 before and needed to move to the new release for two important reasons: MySQL 5.0 and PHP 5. Apache 2.2 was a nice bonus, as well as a new kernel and a host of new software.

Our plan was carefully laid out and we followed it pretty religiously. All in all, everything went well, except the MySQL 5.0 upgrade. I didn't do my homework before the upgrade and so I managed to mess it up, although when using an Entreprise OS I would have assumed the process would have been less painful (as it was for PHP4 -> PHP5). It was my fault.

But it's done. Finished. All our software is up-to-date, and we have no other planned upgrades. Despite the small glitches here-and-there, the upgrade was pretty painless and we didn't have any outages (except a few minutes while MySQL was being naughty). SLES is truly a solid product, and having never worked with SuSE until I started with working at the Foundation, it has earned my respect.

So now what? Oh yeah, now I can get back to all the bugs that are waiting for me in bugzilla - maybe even those that I CANTBEARSED about ;) (someone needs to define CANTBEARSED in Wikipedia...)

Eclipse is Us! Time for Junior Jobs?

I'll jump on the "Eclipse is You" bandwagon, but I'll propose a new theme: can Eclipse be Us instead of just You?

The common message I've been reading in the past few posts is that "You don't do enough". You committers don't do enough, You users don't do enough and You foundation people don't do enough. It's a given -- everyone has limited time, resources and budget, but has endless amounts of work and open bugs. But despite all the discussion, no one has brought any real solutions to the table, other than "You should do more", which is pretty open-ended.

Enter bug 168513, where Frederic suggest Eclipse Committers look at creating Junior Jobs, something KDE already does. Here's the paraphrased concept:
"I'm an Eclipse committer and I think Bug 123456 is pretty simple to do.
Just a (simple, small thing) which uses a couple of calls from (this API).
So simple that it'd fit very well somebody who'd like to start doing
Eclipse development. So simple that I don't want to spend my time on
it, doing the actual coding and testing and all that, and if I'd have
to it'd be probably better spent time if I helped somebody new to do it."
Looking at KDE's bugzilla, there are 137 RESOLVED Junior Jobs bugs.

Eclipse's Bugzilla already has a helpwanted keyword, where committers flag a bug to indicate that they're looking for help from the community. Could another flag, called juniorjobs, or JJ, be created, so that committers can clearly identify small, simple bugs/features that could be easily fixed by the community, or a budding junior developer? Doing so could perhaps liberate some time for the committers to work on bigger stuff.

A perfect example of a JJ bug would be this one: bug 159459.

What do you think? Post your comments on bug 168513.

Thursday, December 14, 2006

Mylar is on fire!

Look who just showed up in #5 spot on the downloads page? Mylar. It's hot - really hot. We'll have to buy more servers just to keep up with the popularity of this project! Well, maybe not, but scoring 5th spot on the top-10 popular Eclipse.org projects is amazing. Congrats to Mik and his team for creating a project that Wayne just can't get enough of. Nathan has been raving about Mylar too.

I haven't had a chance to try Mylar out myself, but I will ...

Thursday, December 07, 2006

Node5 has arrived!

Yesterday we received 'node5' -- a fifth front-end processing node for our cluster. It's essentially identical to the four other nodes: HP Integrity rx2620 with two Itanium2 1.6GHz CPUs, 8GB of RAM and three 146GB SCSI disks connected to a 6420 RAID controller. The server has two Gigabit Ethernet ports, and we added a third Gigabit NIC for our tape backup network.

The front-end cluster handles all the incoming web, CVS, rsync and e-mail connections for all our sites. Load is distributed to the nodes by a Cisco LocalDirector, which can automatically detect a broken node and put it out-of-service. The nodes connect to a pair of backend IBM eServer P5 550 servers for database, ldap authentication and shared storage services. This setup allows us to scale horizontally by adding more nodes as CPU load increases, which is exactly why 'node5' is here. Each node is capable of handling over 600 simultaneous web clients -- and at the same time, a few RSYNC connections and dozens of CVS connections between the dozens of e-mails per minute sent to/from our servers.

We're waiting for a few parts, but Node5 is expected to be in service as early as December 15. Break it in gently ;-)