[mythtv-users] `Mythbackend leaks memory` continued

Alex Butcher mythlist at assursys.co.uk
Wed Aug 13 11:53:12 UTC 2008


On Wed, 13 Aug 2008, Udo van den Heuvel wrote:

> Alex Butcher wrote:
>>> Hmm...
>>> More work needs to be done to tickets like
>>> http://svn.mythtv.org/trac/ticket/5545.
>>
>> A thought occurs to me; has anyone taken into account that MySQL clients'
>> memory usage will vary? I seem to recall reading elsewhere that the MySQL
>> client library has its own internal caching/garbage collection that can be
>> mistaken for memory leaks by the naïve.
>
> Then why would memory consumption keep growing in an almost straight
> line on various systems?
> Garbage collection would assume some process that frees up memory again.

Yup, I'll concede that would appear to be a contradictory piece of evidence.

Conceivably, though, the garbage collection might run only when the host is
otherwise idle. And if it (almost) never is, it (almost) never runs...

> This contradicts the observation of 'continuous' growth:

I've set up a similar cronjob on my machine (FC8, ATrpms, P4 2.53GHz, 768MB
RAM, combined FE/BE, ~400GB recordings, 1xNova-T, 1xNova-T-500, 2 virtual
tuners on every physical tuner, using EIT on the Nova-T, not on the
Nova-T-500) to see if I can observe the same behaviour.

So far, not: for a mythbackend process that's been running for just over 215
hours, it had a VSZ of 339452KB and an RSS of 29984KB.

> Kind regards,
> Udo

Best Regards,
Alex.


More information about the mythtv-users mailing list