Error: Could not shut down filedescriptor...

[Henrique de Moraes Holschuh]

On Tue, 22 Jul 2003, Erik Steffl wrote:
> Henrique de Moraes Holschuh wrote:
> >On Thu, 17 Jul 2003, Rob Siemborski wrote:
> >>On Thu, 17 Jul 2003, Erik Steffl wrote:
> >>>Jul 17 08:56:43 jojda cyrus/imapd[7560]: lock_reopen: blocked, sleeping
> >>>for 0 on interval 0 (15, /var/lib/cyrus/user/e/erik.seen)
> >>>Jul 17 08:56:43 jojda cyrus/imapd[7560]: lock_reopen: blocked, sleeping
> >>>for 1 on interval 1 (15, /var/lib/cyrus/user/e/erik.seen)
> >
> >You're getting lock contention in Berkeley DB, and the no-deadlock stuff I
> >put in Debian is doing exponential backoffs.

I need more coffee.  Debian uses skiplist for the SEEN db, so it is a lock
contention in skiplist.  I am not sure this is good at all, it is the first
I've heard of it.

Although I did not deploy 2.1.14 to production here yet, so it might have
something to do with changes in 2.1.14's skiplist I did not notice in time.

>   I suspect that that might be a problem, I just noticed that it checks 

Well, if it tries to do so at the same time, a lock contention will happen,
but it should be a rather fast one...  however, it appears that by the time
cyrus imapd tries again, the other instance has moved on to another mailbox,
and it gets contention again, and again, and again...

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh