[Dovecot] dovecot-auth: Too many open files

[r.lang at scan-plus.de]

Hi,

we had recently a similar problem, having the following error
messages:

dovecot: imap-login: socket() failed: Too many open files
dovecot: imap-login: proxy(xxxx): \ connect(x.x.x.x), 143) failed: \
Too many open files

Since the 'ulimit -n' is process-dependent we set it to 'unlimited'
in the '/etc/init.d/dovecot' script. This helps, we've got no such
errors since.

Greetz,
Ruben


dovecot-bounces+r.lang=scan-plus.de at dovecot.org schrieb am 23.10.2007 
19:37:27:

> IIRC setting "login_process_size: 64" fixes this.
> 
> p at rick
> 
> 
> * Joe Allesi -X (joallesi - Coyote Creek Consulting at Cisco) 
> <joallesi at cisco.com>:
> > All,
> > 
> > [version: dovecot-0.99.11-4.EL4.src.rpm]
> > 
> > We recently experienced an issue that prevented all new IMAP logins 
from
> > occurring. Although it appears that it was due to running out of
> > available system file descriptors, I'm still not sure what the true 
root
> > cause was as I can't replicate the same error in our test environment.
> > The system file descriptor max was set at (per `cat
> > /proc/sys/fs/file-nr`) 380081 , and we're averaging around 7,000 in 
use.
> > It appears that the main issue was dovecot-auth, so does this appear 
to
> > fall in line with the known PAM bug in this version?
> > 
> > Auth config:
> > auth_userdb = passwd
> > auth_passdb = pam
> > 
> > Checked maillogs (cleansed):
> > 
> > Oct 16 11:48:05 host dovecot-auth: PAM: pipe() failed: Too many open
> > files
> > Oct 16 04:48:06 host imap-login: Disconnected [::ffff:<internal_IP>]
> > Oct 16 11:48:11 host dovecot-auth: PAM: pipe() failed: Too many open
> > files
> > Oct 16 04:48:11 host imap-login: Aborted login [::ffff:<internal_IP>]
> > Oct 16 04:48:11 host dovecot-auth: PAM unable to
> > dlopen(/lib/security/pam_nologin.so)
> > Oct 16 04:48:11 host dovecot-auth: PAM [dlerror:
> > /lib/security/pam_nologin.so: cannot open shared object file: Too many
> > open files]
> > Oct 16 04:48:11 host dovecot-auth: PAM adding faulty module:
> > /lib/security/pam_nologin.so
> > Oct 16 04:48:11 host dovecot-auth: PAM unable to
> > dlopen(/lib/security/pam_stack.so)
> > Oct 16 04:48:11 host dovecot-auth: PAM [dlerror:
> > /lib/security/pam_stack.so: cannot open shared object file: Too many
> > open files]
> > 
> > So, during the event I performed an strace against the dovecot-auth
> > process and noticed the following error as well:
> > 
> > accept(3, 0xbfe05a50, [2])              = -1 EMFILE (Too many open
> > files)
> > gettimeofday({1192538310, 41972}, NULL) = 0
> > poll([{fd=3, events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL,
> > revents=POLLIN}, {fd=0, 
events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL},
> > {fd=18, events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=10,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=7,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=23,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=8,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=11,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=12,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=13,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=14,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=19,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=16,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=17,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=6,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=15,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=20,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=21,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=9,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=22,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=24,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=25,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=26,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=27,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=28,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=34,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=33,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=29,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=37,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=30,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=31,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, {fd=41,
> > events=POLLIN|POLLPRI|POLLERR|POLLHUP|POLLNVAL}, ...], 1020, 1585) = 1
> > gettimeofday({1192538310, 99354}, {480, 0}) = 0
> > 
> > Then I checked for the current network connections:
> > 
> > netstat | grep imap | wc -l
> > 42
> > 
> > Then used lsof to check for open files:
> > 
> > while true ; do lsof | awk '{ print $3 }' | wc -l ; sleep 4 ; done
> > 21728
> > 21668
> > (restarted dovecot)
> > 1838
> > 1953
> > 1864
> > 2066
> > 2018
> > 2020
> > 2036
> > 2189
> > 2661
> > 2558
> > 2490
> > (cont @2,000)
> > 
> > Thanks!
> > 
> > Joe Allesi
> 
> -- 
> state of mind
> Agentur für Kommunikation, Design und Softwareentwicklung
> 
> Patrick Koetter            Tel: 089 45227227
> Echinger Strasse 3         Fax: 089 45227226
> 85386 Eching               Web: http://www.state-of-mind.de
> 
> Amtsgericht München        Partnerschaftsregister PR 563