Upstart gssd.conf: stopping prematurely with kerberized NFS4 mounts
Bug #569094 reported by
Valentijn Sessink
This bug affects 2 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
nfs-utils (Ubuntu) |
Fix Released
|
Medium
|
Unassigned | ||
Lucid |
Won't Fix
|
Medium
|
Unassigned |
Bug Description
/etc/init/gssd.conf states:
stop on (stopping portmap or runlevel [06])
When going to runlevel 0 or 6, this results in unusable (and even hanging) Kerberized NFS4 mounts - i.e. if there's any process still using anything on nfs4, it can't continue because gssd isn't there anymore.
The right moment to stop gssd would probably be after unmounting any NFS4 mounts.
Changed in nfs-utils (Ubuntu Lucid): | |
assignee: | Steve Langasek (vorlon) → nobody |
milestone: | lucid-updates → none |
Changed in nfs-utils (Ubuntu): | |
assignee: | Steve Langasek (vorlon) → nobody |
To post a comment you must log in.
Yes, now that the sysvinit package has been fixed to not kill processes belonging to upstart jobs, the main reason to kill the process ourselves goes away.
So I was going to mark this as a candidate for SRU in lucid.... however, because rpc.gssd is located in /usr/sbin, leaving it running could prevent being able to cleanly unmount /usr on shutdown. So we need to be able to stop it after /etc/rc6. d/S31umountnfs. sh runs, but before /etc/rc6. d/S40umountfs runs, and there's currently no way to do that with an upstart job in lucid. We'll have to figure out a way to fix this for maverick.
In the meantime, users affected by this should edit /etc/init/gssd.conf to be 'stop on stopping portmap'.