After a LONG session on the terminals:
I have no f***ing clue what it is and I need a workaround. As far as I see, it it mostly postfix's trivial-rewrite programm that dies.
I've seen two differences in you my.cnf setup: I did not have skip-bdb and skip-innodb . "skip-innodb" is working - I have no InnoDB here. But as soon as I say "skip-bdb" mysql does not restart! I have no idea why.
I've tried to give MYSQL even more resources than you said, but to no avail. We NEVER have more that 0.1 server load when the problems occur. It is very often at the very first contact of a mailserver coming via SMTP.
Next, I've avoided to use postfix's proxy command proxy:mysql:/etc/postfix.. but mysql:/etc/postfix instead so the commands do not get queued but can hit the mysqlserver with more threads (why not, we have 4 cpus?). No difference.
What I'd suggest is:
Postfix uses no longer MySQL to ask for data, but flat files.
Is there a chance that ISPConfig produces those flat file on demand? I think it was like this back in IPSConfig 2...?
I have restarted the whole server just in case..
As long as we cannot identify the error (and it does definitely not look like), I really need a workaround. My Clients are beginning to realize something is wrong.