Public bug reported:
Binary package hint: pm-utils
I have a variety of machines which mount user home directories over nfs.
When people attempt to hibernate them, the hibernation process hangs
horribly.
Having looked at this, the main problem was that pm-hibernate unhooks
the network first (in /usr/lib/pm-utils/sleep.d/55NetworkManager) but
expects to then be able to call "sync". The latter is clearly never
going to work if there are nfs mounted filesystems and no network.
Problem is, you can't do what many people suggest in other related bug-
reports, of unmounting the nfs filesystems if they're in active use,
with processes having open file descriptors on them. Or rather you
_could_ do a forced unmount, but it wouldn't be desirable.
So I wondered why it was necessary to tell NetworkManager to shut down
the network. I disabled the 55NetworkManager script, and suddenly
hibernation worked, and restarting afterwards worked fine, with no ill
effects.
I've only checked this a couple of times so far; and only over periods
of up to 30 minutes.
Also, my machines are given static dhcp leases (ie not from a pool), so
never change IP address. I'm not entirely convinced that dynamic dhcp
allocation should be problematic though; it should surely be no
different waking up from hibernation and finding your lease has expired,
to just finding your lease has expired ...
I'm wondering what the motivation for the 55NetworkManager sleep.d
script is, as I assume it wasn't just added for fun 8-)
** Affects: pm-utils (Ubuntu)
Importance: Undecided
Status: New
--
pm-hibernate hangs if nfs filesystems mounted
https://bugs.launchpad.net/bugs/681829
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
