Comment 32 for bug 616287

@Clint Byrum - bug 603363 seems to talk about an issue where sshd is not properly being stopped, this was fixed for maverick, but apparently not for lucid. I ran lsof in the umountroot script, so right before the root fs should be remounted read only. I don't think anything gets killed in there between my lsof and the actual remount. The lsof consistenly shows 2 things running: init and sshd. Modifying the ssh.conf file to stop sshd on anything but runlevel 2345 solves that issue. However, I think we're also still running into bug 188925. It appears that right after a libc6 upgrade it fails to properly remount root because it's busy. This leaves orphaned inodes and possibly some other mess. This doesn't happen on a completely clean install without updates installed. All of this goes away after the initial reboot after updating - so this has to be something related to the updates.

As for if this is completely related to this bug, I'm not sure. I'm just seeing this behaviour and I'm still not 100% sure 'what to blame'. Nonetheless, having a dirty filesystem on a new install (a production server even!) right after doing some initial updates is kind of not making me feel very confident about this ;-)