On Friday 15 November 2013 09:04:11 Mike Bonnet wrote:
On 11/15/13 5:48 AM, Didier Fabert wrote:
> Hi all,
>
> I have strange error since last week on my koji. All regen-repo tasks are
> failed with error LockError: [Errno 37] No locks available
>
> Koji works perfectly since January month. Directory /mnt/koji is a nfs
> mount but i can touch a new file on it. Database is ok too.
>
> I have two builders, and error appear on both.
>
> Any idea ?
>
koji.LockError is only raised in two places, both around fnctl.lockf():
http://docs.python.org/release/2.6.8/library/fcntl.html#fcntl.lockf
The fcntl manpage says:
ENOLCK Too many segment locks open, lock table is full, or a remote
locking protocol failed (e.g., locking over NFS).
Are you mounting /mnt/koji over nfs? Maybe the nfslock service died?
Is something else holding a lot of locks?
Yes, you're right, nfslock was dead on NFS server (el5)
After restarting nfslock service on my NFS server, all is OK.
Thanks a lot Mike to pointing me to the right direction.
--
Didier