[Dovecot] mail_max_userip_connections exceeded.
Hi
I'm using Dovecot version 1:1.2.15-7 installed on Debian Squeeze via apt-get..
I have this error in the logs. /var/log/mail.log.1:2490:Jan 19 12:02:55 mail dovecot: imap-login: Maximum number of connections from user+IP exceeded (mail_max_userip_connections): user=user@example.com, method=PLAIN, rip=127.0.0.1, secured
I never changed this from the default 10. When I googled this error there was a thread on this list from May 2011 that indicated one would need one connection per user per subscribed folder. However, I know that user doesn't have 10 folders, let alone 10 subscribed folders! I can increase, it but it's not going to scale well. And there are people on this list with many 1000x users than I have - so how do they deal with that?
127.0.0.1 is obviously webmail (IMP5).
So, how/why am I seeing this, and should I be concerned?
Simon
On 01/20/2012 06:06 PM, Simon Brereton wrote:
I have this error in the logs. /var/log/mail.log.1:2490:Jan 19 12:02:55 mail dovecot: imap-login: Maximum number of connections from user+IP exceeded (mail_max_userip_connections): user=user@example.com, method=PLAIN, rip=127.0.0.1, secured
I never changed this from the default 10. When I googled this error there was a thread on this list from May 2011 that indicated one would need one connection per user per subscribed folder. However, I know that user doesn't have 10 folders, let alone 10 subscribed folders! I can increase, it but it's not going to scale well. And there are people on this list with many 1000x users than I have - so how do they deal with that?
127.0.0.1 is obviously webmail (IMP5).
So, how/why am I seeing this, and should I be concerned?
Well, it really does look like IMP is using more than 10 connections at the same time. Or perhaps some of the existing connections are just hanging for some reason after IMP already discarded them, such as maybe a very long running SEARCH command was started and IMP then gave up. You could look at the process list (with verbose_proctitle=yes) and check if the user has other processes hanging at the time when this error is logged.
participants (2)
-
Simon Brereton
-
Timo Sirainen