[Tracker-discuss] PSF tracker causing heavy load

[anatoly techtonik]

On Wed, May 12, 2010 at 6:54 PM, R. David Murray <rdmurray at bitdance.com> wrote:
>
> I've been taking a look, but have so far been unable to identify anything
> triggering the load.  It seems to come in bursts, and when it does, a
> bunch of roundup processes accumulate while the requests are dealt with,
> and the postgres CPU use goes up as well.
>
> I started a script to find the start of a spike so I can try to correlate
> it with events in the logs, but haven't been able to get a responsive
> enough shell to look at the results yet :)

Add timestamps to logs. I guess there is some monitoring tool that
lists time of spikes as well. Is there any? Then map spikes to search
queries at this time. I guess roundup doesn't have multilevel logging
to turn on tracing of search queries and their time profiles, and dump
them to separate file or socket.

> On thing that would at least cut down on the load on roundup (based
> on my perusing of the access logs) would be to block the msnbot temporarily.
-- 
anatoly t.