hi,
we (the debian maintainers for zabbix) have received a
bugreport by a user which is somehow related to how
zabbix-server and zabbix-agent[d] handle logfiles. Most
distributions provide some kind of logrotate facility,
so does Debian.
This means logrotate takes care of rotating logfiles for
the different Services (like apache etc.) (with renaming,
gzipping, restarting the service after rotation and so
on ..)
Now, zabbix-server renames its logfiles to .old if the
logfile size is greater than 1024*1024 (and overwrites
this file then). This leads to one problem: If a user
wants to debug his Server over a few days with a high
debug level, some logging information may be lost due to
the .old files getting overwritten.
So, this leads to one question: should zabbix, if caring
about rotating its logfiles, rotate them in some kind of
safe way (like, zabbix-server.log.1 zabbix-server.log.2)
and take care about compressing them (like logrotate does)
or simply should not care about this task and leave it up
to the distributors (by using logrotate) to care about
log-rotation (i dont know how *BSD or other *nixes like
Solaris handle this).
we (the debian maintainers for zabbix) have received a
bugreport by a user which is somehow related to how
zabbix-server and zabbix-agent[d] handle logfiles. Most
distributions provide some kind of logrotate facility,
so does Debian.
This means logrotate takes care of rotating logfiles for
the different Services (like apache etc.) (with renaming,
gzipping, restarting the service after rotation and so
on ..)
Now, zabbix-server renames its logfiles to .old if the
logfile size is greater than 1024*1024 (and overwrites
this file then). This leads to one problem: If a user
wants to debug his Server over a few days with a high
debug level, some logging information may be lost due to
the .old files getting overwritten.
So, this leads to one question: should zabbix, if caring
about rotating its logfiles, rotate them in some kind of
safe way (like, zabbix-server.log.1 zabbix-server.log.2)
and take care about compressing them (like logrotate does)
or simply should not care about this task and leave it up
to the distributors (by using logrotate) to care about
log-rotation (i dont know how *BSD or other *nixes like
Solaris handle this).
, sorry:
Comment