Re: [cobalt-users] RAQ4 slowing down with high processor load.

[Lauchlin <lauchlin_list@xxxxxxxxxxxx>]

parseReport.pl is the script that creates the stats the the admin 
interface.  If you set the size of virtual website really big (our home 
site is 15GB) log rotate won't rotate the logs until they are 10% of 
the size of the sites quota.  This means that after a few years of 
operation you could potential have some really really big logs, which 
of course start to be a problem with parseReport.pl.  The solution I 
fond was to modify the way logrotate and the parseReport.pl work by 
first reducing the size that logrotate uses down to 1% or something 
similarly small then secondly deleting a few lines in parseReport.pl 
that create the internal admin statistics.  I can post more details if 
you wish but maybe that will head you on the correct track.

Cheers,

Lauchlin


On 02/02/2004, at 10:10 AM, Neil James wrote:

> Hi
>
> we have a RAQ4 128MB which in the last few weeks has been giving 
> problems.
> It's a standard machine except we have webalizer installed. Nothing 
> new has
> been installed or patched recently.  The box has been running fine for 
> almost 3
> years.
>
> It only has about 10 active sites which get 300-400 visitors a day at 
> most
> although we did have one quite busy site (>6000 visitors per day) 
> which I moved
> to a server of it's own about a week ago.
>
> We've started getting daily error messages (timed > 0400) that it was 
> very low
> on memory. Sometimes we'd get 'postgresql server is not responding'.
> Eventually we would get a message that the processor load was very 
> high.
>
> Going to the control panel would bring up the left frame but 'internal 
> server error'
> messages for the main and status frames. Rebooting it cleared it but 
> after a
> while I did notice that we were not getting access stats since the 
> problem
> started.
>
> We also had error messages like this:
>
> errors occured while rotating /var/log/httpd/access
>
> Out of memory!
> Callback called exit at /usr/local/sbin/split_logs line 7.
> END failed--cleanup aborted at /usr/local/sbin/split_logs line 48.
> Callback called exit at /usr/local/sbin/split_logs line 48.
> END failed--cleanup aborted at /usr/local/sbin/split_logs line 31.
> Callback called exit at /usr/local/sbin/split_logs line 31.
> END failed--cleanup aborted at /usr/local/sbin/split_logs line 58.
> Callback called exit at /usr/local/sbin/split_logs line 58.
> BEGIN failed--compilation aborted at /usr/local/sbin/split_logs line 7.
> error running prerotate script -- leaving old log in place
>
> The host company tech support suggested running a daily script to rm
> web.cache files from all sites, which didn't help. I tried deleting 
> access logs too
> but no change.
>
> Running top showed a process called parseReport.pl using  (typically) 
> 168M of
> memory that had been running for 13 hours! I'm not sure what this 
> process is but
> think it might be associated with webalizer.
>
> Can anyone tell me if this process is webalizer and if so what is 
> causing it to
> choke? Is it a build up of log sizes somewhere?
>
> If so how do I reduce the size? Even more important how do I find them?
>
> My expertise is limited to running the Cobalt control panel so any 
> pointers would
> be much appreciated.
>
> Thanks in advance
>
> Neil
>
> _____________________________________
> cobalt-users mailing list
> cobalt-users@xxxxxxxxxxxxxxx
> To subscribe/unsubscribe, or to SEARCH THE ARCHIVES, go to:
> http://list.cobalt.com/mailman/listinfo/cobalt-users
--
Lauchlin Wilkinson, BComp
Technical Manager
Internet Tasmania Pty. Ltd.