[rsyslog-notify] Forum Thread: Postfix reliable setup - (Mode 'post')

[noreply at adiscon.com]

User: sbmk 
Forumlink: http://kb.monitorware.com/viewtopic.php?p=24856#p24856

Message: 
----------
Hi,
I'm trying to build a new mail architecture, with reporting capabilities.
So I decided to use rsyslog with a mysql database. 
As it's only a testing setup, I just added the "@@" line on my clients,
keept the defaut filesystem directives too.
It was running flawlessy but I had a bad suprise : when my rsyslog server
became full (the filesystem containing the database), the postfix clients
quickly stopped working and didn't deliver anything more. When I commented
the new line and restarted rsyslog, everything was fine again.
Of course, I was using TCP and "$ActionResumeRetryCount -1" in order to
keep as reliable as possible, but it seems it can be dangerous when things
go bad on the server side for a long time.

Switching to UDP looks like a trivial solution, but I'd like to stay on the
"acknowledged side".

I think I'll keep some local logfiles, just in case of, but I need a setup
which could survive a log server crash without compromsing the clients.

What's the beast approach to get something reliable as long as the serveur
is running, but "forgetting" if it fails ?

Regards