Device issues with disk images ...
Then … a few weeks ago, I rebooted my Linode for an upgrade, and suddenly the swap and home partition wouldn't mount any more. I poked around a bit, and saw that in the configuration profile I was using that the devices were listed as /dev/udba, /dev/udbb, and /dev/udbc. So, I changed my fstab accordingly, rebooted, and everything was fine.
Now, just a few minutes ago, my Linode got rebooted because of a kernel panic on the host machine, and once again it wouldn't mount the swap and home partitions. So, I went back, looked at the configuration profile, and the devices are still listed as ubda, etc. However … poking around on the Linode, I see that the devices are /dev/udb/0 again.
Am I doing something wrong here? Going crazy?
I'm running Gentoo, if that makes any difference.
6 Replies
Or, if you're on host40, it could be because during the maintenance yesterday they reverted the host kernel to an older version. But that doesn't make sense if you rebooted your VDS and then it didn't work, so I'll guess it's a problem with the VDS kernel, not the host kernel.
caker
Linode Staff
The Configuration Profile on the website has always read "/dev/ubd[a-h]" even though Gentoo users need to use /dev/ubd/[0-7].
-Chris
@caker:
I had a report of this a few days ago with a guy I helped out in the IRC channel. We haven't changed anything, and I can't think of anything that would cause your devfs to load/not to load.
The Configuration Profile on the website has always read "/dev/ubd[a-h]" even though Gentoo users need to use /dev/ubd/[0-7].
-Chris
Honestly, I'd never paid attention to what the images were listed as on the configuration profile page until things weren't working the first time. I just knew that I couldn't mount the images … mount complained that the devices didn't exist … and then when I checked to make sure that the images were still listed, I noticed the discrepancy between what was listed their and what was listed in my fstab.
It's really interesting that it booted and succesfully mounted the images using the /dev/dba[a-h] terminology. I'm pretty sure that during that time period I rebooted at least once for a glibc update.
I'm still in the middle of upgrading, so I guess I'll see if things magically change again. I am not going to start my build again from scratch just to test this theory though…
btw, turning off RCDEVICETARBALL in /etc/conf.d/rc seems to cleanup the crud in /dev
