Yesterday I noticed that my 12.1 server was out of disk space. Nothing visible, but df said the server was full when du said it was not.
A friend of mine said he had the same problem, that logrotate did not restart the application after rotating logs and that so the removed files where still open “on the background”, so still using room.
In fact, rebooting the server fixed the problem (df=du).
If the diagnostic is good, it may be Inn that caused it because I have an unknown error with it (not to be discussed here).
Do you know if there is a permanent solution to such problem?
holler if neither of those help–by the way, if you have a runaway error
filling logs you are gonna have to hack your logrotate to deal with it
(like: compress sooner and move the old logs out faster, for example)
and, maybe you can ‘teach’ Inn to send errors to the bit bucket (if you
are gonna ignore them anyway), see man inn)
On 2012-09-13 16:25, Ken Schneider wrote:
> On 09/13/12 09:58, Ken Schneider pecked at the keyboard and wrote:
>> Or better still, find out what is filling the logs and correct the problem.
> In this case find out why logrotate is not working properly and get it fixed. logrotate should
> not be keeping previous files open. Perhaps a bug report is in order.
Notice that often logrotate restarts the writing process to avoid that precisely. See, for
example, “/etc/logrotate.d/syslog”, where there is a line that does “/etc/init.d/syslog reload
Cheers / Saludos,
Carlos E. R.
(from 12.1 x86_64 “Asparagus” at Telcontar)