On Thu, 2006-11-09 at 17:09 +0100, Gábor Lénárt wrote:
Hello again,
Now found more and more strange messages, like:
pop3-login: Can't connect to auth server at default: Resource temporarily unavailable
This means that dovecot-auth is busy and couldn't respond in time.
auth(default): LDAP: ldap_result() failed: Can't contact LDAP server
auth(default): ldap(user,xxx.xxx.xxx.xxx): ldap_bind() failed: Server is unwilling to perform
Well, this sounds like it really is down.
And the most bizarre stuff:
auth(default): ldap_bind((null)) failed: Can't contact LDAP server
This is strange. I tried to figure out a reason why it could happen, but couldn't.. I added anyway an extra assert and changed one error string a bit, so if this happens to you again in rc14 I could maybe debug it further. Especially if it crashes, gdb backtrace would be helpful.
My problem that dovecot pop3 daemon dies after a while (after some hours): when reach that point, POP3 login no longer works, after issuing POP3 command "PASS", it just hangs. Stopping and restarting dovecot cures it for some hours at least. The strange that these error log messages are produced all the time, but not always cause the hang of the WHOLE service.
I guess the connection to LDAP server dies somehow. Does OpenLDAP have some timeouts for connections? That might cause it I guess..
If someone have ANY information, please help me. We used tpop3d before, and that also had similar problems (eg hanging the whole service after a while). That LDAP server is OpenLDAP from Ubuntu 6.06 LTS on AMD64 (2.2.26-5ubuntu2.1). You may think that if both of tpop3d and dovecot fails similary it should be the problem of the LDAP server, however other services (web interface, postfix) does work without a single problem since a month or so. Tried with v2 and v3 LDAP protocol as well. I ran out of ideas ;-(
I think the difference is that Dovecot keeps the connection open all the time, while web/postfix keep reconnecting to it constantly.