toofishes.net

Slicehost kernel upgrade

I decided to reboot my slice today after more than 70 days of uptime. The reason was to try out the new kernel they made available due to the release of Ubuntu Karmic Koala a few days ago. If only it didn’t take a release of Ubuntu to get more up-to-date software, but I’ll fight that argument another day.

We all know Arch tends to be bleeding edge, so it is good to get back to running a more recent kernel. I was previously running a version labeled 2.6.24-24.55; I’m now on 2.6.31-302-rs. I don’t really understand the crazy version numbers, but clearly one is a heck of a lot more recent.

So the awesome results of this upgrade? Nothing seemed to break, which means things went well. However, some super important words of warning: when you do the kernel upgrade via the SliceManager, the slice is not nicely rebooted. It is immediately killed; why on earth they decided to go with xm destroy instead of xm shutdown I have no idea. I’d recommend you shut down services on your own before you do the upgrade or you will get beautiful messages in your database logs telling you how stupid you were to not cleanly shutdown like I did.

Oh, and a possible bonus after the upgrade- all the sudden my 256 MB slice got turned into a 321 MB slice. We’ll see how long that lasts. :)

EDIT: I did encounter one small bit of fallout- the new kernel doesn’t include the xt_recent module to replace the older ipt_recent module, so my iptables rules failed to load because I have some fancy SSH rate limiting rules. For now, I sent them a support email letting them know and loaded all of the other rules that don’t need the recent iptables module.

Tags

See Also