apache2/rotatelogs2 - problems with disappearing processes

[der bastler]

I use pipes and rotatelogs2 to collect the logs of two virtual hosts. When I start apache2, all six rotatelogs2 processes (error/access for two hosts and main server) and five apache2 child processes (user apache) are started.

But after some time -- between half a day and two days -- the rotatelogs processes plus the main apache process are gone. Webserver is still serving pages, but does not write logs anymore. Stopping the apache2 via initscript gives an error about its pid and the remaining child processes must be killed manually. After a new start of the initscript everything's fine again... for a while.

Couldn't find the problem here or elsewhere. Do I have to abandon the rotatelogs strategy?
_________________
Tempus fugit.
@frank@troet.cafe

[der bastler]

Well, now I'm using ordinary files and logrotate.

In the last days the apache processes were pretty stable: some children, one root process.

In my log files I found setuid issues, so I reviewed /etc/limits and /etc/security/limits.conf to make sure nproc is set to a bigger value. Funny, thought I did it beforehand.

Confused[1], but up and running. *shrugs*


--

[1] The otherwise excellent Gentoo documentation is a little bit fuzzy about /etc/limits, /etc/security/limits.conf, when to use what and requirements of mainstream server applications...
_________________
Tempus fugit.
@frank@troet.cafe