Hi all,
perhaps someone has some pointers or experience for me in the following (argh!) case: my webmail suddenly does not work anymore but pop/smtp/imap works fine.
It's a long post, but I didn't want to bother anyone without doing some serious research first.
Everything worked great until one day, squirrelmail starts to have a problem - when you send an email, it IS sent but squirrelmail gets stuck in the "sending mail" routine (and eventually times out). The sent email is not placed in the sent mail folder. Roundcube mail works fine at this time, and does place sent emails in the sent mail.
Then, a day later, squirrelmail AND roundcube both stop functioning and give the following error:
It is not possible to log in through webmail anymore.
I restarted MailScanner (which restarts exim), restarted Dovecot, restarted MySQLd, restartde httpd and upgraded Squirrelmail but to no avail.
Log entries that may give some clues:
In my nightly logwatch I received this about 10 times:
In /var/log/message about 8 of these (but not exactly at the time of testing, so maybe not related!):
Google suggested that this could be due to a full disk, or due to SElinux issue)
In the httpd log of one of the domains where I tried to log in:
In the Dovecot log:
only the webmail clients connect with localhost. Except for this, I only get succesful logins through POP3/IMAP in the Dovecot log.
Most importantly (considering the first error), my /var/ partition probably did fill up momentarily, because one client received tens of thousands of emails from one of his suppliers due to a mailing problem (...). By the time I received notification there was free diskspace again.
Jeff mentioned in another post that a full disk may result in corrupt files. Since the var partition only hold logs and incoming mail (until it is delivered to the inboxes in /home), I wonder whether anything important to the system could get hit.
Lastly, I tried to connect to localhost port 143 through telnet and was succesful. When I run a remote Roundcube (or pop/imap) everything is perfectly reachable and functional.
So the problem somehow lies in the link between the webfront-end and Dovecot. This involved php, apache, the webinterface scripts, transport to localhost and dovecot. Dovecot works, the webinterface scripts I updated without change.
Hope you want to give me some suggestions - thank you for any!
Harro
perhaps someone has some pointers or experience for me in the following (argh!) case: my webmail suddenly does not work anymore but pop/smtp/imap works fine.
It's a long post, but I didn't want to bother anyone without doing some serious research first.
Everything worked great until one day, squirrelmail starts to have a problem - when you send an email, it IS sent but squirrelmail gets stuck in the "sending mail" routine (and eventually times out). The sent email is not placed in the sent mail folder. Roundcube mail works fine at this time, and does place sent emails in the sent mail.
Then, a day later, squirrelmail AND roundcube both stop functioning and give the following error:
Code:
Error connecting to IMAP server: localhost.
115 : Operation now in progressIt is not possible to log in through webmail anymore.
I restarted MailScanner (which restarts exim), restarted Dovecot, restarted MySQLd, restartde httpd and upgraded Squirrelmail but to no avail.
Log entries that may give some clues:
In my nightly logwatch I received this about 10 times:
Code:
Cannot mkdir /var/spool/MailScanner/incoming/8354/1IzO2Q-0002yp-Kg, No space left on device : 1 Time(s)
Cannot create temporary Work Dir /var/spool/MailScanner/incoming/11966. Are the permissions and ownership of /var/spool/MailScanner/incoming correct? : 1 Time(s)In /var/log/message about 8 of these (but not exactly at the time of testing, so maybe not related!):
Code:
bitmap kernel: post_create: setxattr failed, rc=122 (dev=md6 ino=4574097)In the httpd log of one of the domains where I tried to log in:
Code:
[error] [client 154.98.50.228] PHP Notice: Could not connect to mail.domain.nl at port 143: Operation now in progress in /var/www/html/roundcubemail-0.1beta2/program/include/bugs.inc on line 80, referer: http://www.domain2.nl/webmail/In the Dovecot log:
Code:
dovecot: Info: imap-login: Disconnected: rip=127.0.0.1, lip=127.0.0.1, securedMost importantly (considering the first error), my /var/ partition probably did fill up momentarily, because one client received tens of thousands of emails from one of his suppliers due to a mailing problem (...). By the time I received notification there was free diskspace again.
Jeff mentioned in another post that a full disk may result in corrupt files. Since the var partition only hold logs and incoming mail (until it is delivered to the inboxes in /home), I wonder whether anything important to the system could get hit.
Lastly, I tried to connect to localhost port 143 through telnet and was succesful. When I run a remote Roundcube (or pop/imap) everything is perfectly reachable and functional.
So the problem somehow lies in the link between the webfront-end and Dovecot. This involved php, apache, the webinterface scripts, transport to localhost and dovecot. Dovecot works, the webinterface scripts I updated without change.
Hope you want to give me some suggestions - thank you for any!
Harro