Just to add a bit but important information: postfix is run with mailboxes (maildir) on GFS (cluster environment). After some tweaks with the cluster configuration, test with ping_pong gives a performance of 2000locks/s, I wonder if it's enough for postfix with a system of 5000+ user mailboxes...
Zozime --- On Mon, 4/11/11, Wietse Venema <wie...@porcupine.org> wrote: > From: Wietse Venema <wie...@porcupine.org> > Subject: Re: email delivery delay : fatal: shared lock active > To: "Postfix users" <postfix-users@postfix.org> > Date: Monday, April 11, 2011, 4:35 PM > Zozime Rakotondrazafy: > > Hello there, > > > > > > We have been facing this problem for few weeks and > could not find > > the right solution yet... One of our customers' system > is experiencing > > a persistent email delivery delay (either for local or > oustide > > recipients) and after making multiple changes to > postfix settings, > > we keep on having the same error messages in the log: > "fatal: > > shared lock active". > > No, the error message is: > > fatal: shared lock active/9978637B4A > > Where active/9978637B4A is the relative pathname of the > queue file. > > > Mar 30 07:51:30 mailserver postfix/local[24516]: > fatal: shared lock active/9978637B4A: Resource temporarily > unavailable > > Mar 30 07:51:30 mailserver postfix/local[24519]: > 9978637B4A: to=<recipie...@mydomain.com>, > relay=local, delay=0.26, > > Mar 30 07:51:30 mailserver postfix/local[26515]: > 9978637B4A: to=<recipie...@mydomain.com>, > relay=local, delay=0.27, > > Process 26515 got a shared lock, process 24516 got no lock, > and > process 24519 got a shared lock. > > Check your kernel parameters. Perhaps your OS has a > very limited > number of locks available. This may be the result of very > cheap > shared hosting. > > Wietse >