Logrotate: don't seem to be able to restart service (apache)

Swakoo · 01-26-2008, 08:35 AM

Hi guys,

I have in my server, 2 instances of apache 2.0.54.
One is relying on java's Policy Agent, one is not.

Anyway, I am trying to use logrotate to rotate the logs so as to save some space... but I realise the service is not being restarted as the process ID is the same...

this is the logrotate conf I used

Quote:

/var/log/httpd/*log {
daily
rotate 5
missingok
notifempty
sharedscripts
postrotate
apachectl -f conf/withpa.conf stop
apachectl -f conf/httpd.conf restart
apachectl -f conf/withpa.conf start
endscript
nocompress
}

the instance with pa have to be stop and then start, thus so

am I doing it wrongly?

Also, I am thinking of backing up the logs to a remote server using scp on a daily basis... but can't think of a good way to let the script copy the latest log... since the logs are being rotated and then renamed...

any pointers?
thanks!

gilead · 01-26-2008, 01:31 PM

I stopped using logrotate for Apache a while back and now I let Apache rotate the logs itself. From the man page for rotatelogs:

Code:

 CustomLog "|bin/rotatelogs /var/logs/logfile 86400" common

       This  creates the files /var/logs/logfile.nnnn where nnnn is the system time at which the log nominally starts
       (this time will always be a multiple of the rotation time, so you can synchronize cron scripts  with  it).  At
       the end of each rotation time (here after 24 hours) a new log is started.
              
            CustomLog "|bin/rotatelogs -l /var/logs/logfile.%Y.%m.%d 86400" common

Swakoo · 01-26-2008, 01:41 PM

Quote:

Originally Posted by gilead

I stopped using logrotate for Apache a while back and now I let Apache rotate the logs itself. From the man page for rotatelogs:

Code:

 CustomLog "|bin/rotatelogs /var/logs/logfile 86400" common

       This  creates the files /var/logs/logfile.nnnn where nnnn is the system time at which the log nominally starts
       (this time will always be a multiple of the rotation time, so you can synchronize cron scripts  with  it).  At
       the end of each rotation time (here after 24 hours) a new log is started.
              
            CustomLog "|bin/rotatelogs -l /var/logs/logfile.%Y.%m.%d 86400" common

cool i didn't know you could do that.. i'll check it out.. thanks!

Swakoo · 01-28-2008, 12:37 AM

hi guys.. but just thought i carry on from my initial question

if i do logrotate, how do I ensure the service restart? is my way correct?

gilead · 01-28-2008, 01:06 PM

The example in the logrotate man page uses a postrotate option to re-start the daemon:

Code:

"/var/log/httpd/access.log" /var/log/httpd/error.log {
  rotate 5
  www@my.org
  size=100k
  sharedscripts
  postrotate
    /sbin/killall -HUP httpd
  endscript
}

Swakoo · 02-03-2008, 09:24 PM

Quote:

Originally Posted by gilead

The example in the logrotate man page uses a postrotate option to re-start the daemon:

Code:

"/var/log/httpd/access.log" /var/log/httpd/error.log {
  rotate 5
  www@my.org
  size=100k
  sharedscripts
  postrotate
    /sbin/killall -HUP httpd
  endscript
}

yeah but mine is self-compiled.. will it catch it?
Or does the flag -HUP mean that?

gilead · 02-03-2008, 09:31 PM

Your Apache is self-compiled? No problem, mine is too and logrotate works properly here.

Swakoo · 02-04-2008, 06:33 AM

Quote:

Originally Posted by gilead

Your Apache is self-compiled? No problem, mine is too and logrotate works properly here.

actually my concern is if my apache is being restarted properly.
Initially I have the misconception that if i restart it, the process ID will be different.. but apparently not.. so how do I see if apache restarted successfully?

also.. for logrotate.. is it essential for the service to be restarted after rotation?

lastly.. what does the -HUP mean? does it mean that it will restart the process?

gilead · 02-04-2008, 12:50 PM

I need to use the killall on my setup. If you want to see the different effects, try it with and without, but make sure that you hit your web servers with requests before and after so you can see which log files are being used. httpd should be still running, you can use netstat to confirm it:

Code:

netstat -tunape | grep httpd

The HUP signal causes Apache to re-read its config files and re-initialise.

Swakoo · 02-09-2008, 03:47 AM

hi there, i've used the -HUP but to no avail. (/sbin/killall -HUP httpd)
My logs are rotated, but it won't write to the new logfiles, unless i manually restart apache.

I've now write a script to restart apache, and have point logrotate to restart it.. will know tomorrow if it works.
Meanwhile... how do you guys do it? My apache is self-compiled. I am running multiple instances of apache in the same machine, so I have to restart all of them.

thanks!

gilead · 02-09-2008, 01:13 PM

The problem you're describing is the reason I use Apache's rotatelogs utility. If you manage to get it working the way you described earlier, can you post your final config please?