Hi Timo and list members,
Timo Sirainen wrote:
On 1.8.2012, at 23.25, Daniel Parthey wrote:
The error is still the same "config permission denied" shown above? I found that also from my server and added a debug patch, but it hasn't crashed yet. Could you try the attached patch and getting a gdb backtrace from the resulting core file? (Or at least the raw backtrace - getting a core file might be tricky.)
Running command on a four host setup with mailbox+director instance each: /usr/bin/doveadm -c /etc/dovecot-director/dovecot-director.conf -D purge -A
dovecot: doveadm(username@example.org): Error: userdb lookup: connect(/var/run/dovecot/auth-userdb) failed: Permission denied (euid=501(vmail) egid=123(vmail) missing +r perm: /var/run/dovecot/auth-userdb, we're not in group 122(dovecot), dir owned by 0:0 mode=0755)
Ah, so the original patch helped! This is a different error.
srw-rw---- 1 dovecot dovecot 0 2012-07-11 18:35 auth-userdb
Simplest solution now would be to make this world-rw, see the auth-userdb socket configuration in http://wiki2.dovecot.org/LDA#Virtual_users
But I guess this should also be fixed by doveadm-server. Although I don't think this should be happening by default anyway. Maybe this is also solved by the http://hg.dovecot.org/dovecot-2.1/rev/476381017ec7 patch?
I finally found time to update from 2.1.8 to 2.1.10 and change service auth-user db socket to default mode of 0666.
Unfortunately, the issue is still not solved and I did not manage to get a gdb backtrace, since it does not crash or assert. Current configuration of both mailbox and director is attached.
The error "Permission denied" from the mailbox logs is gone, but the director doveadm command: /usr/bin/doveadm -c /etc/dovecot-director/dovecot-director.conf -D purge -A still throws the error message:
doveadm(username@example.org): Error: doveadm server failure
doveadm: Error: Failed to iterate through some users
Any idea what I could do in addition to making /var/run/dovecot/auth-userdb world-rw?