[mythtv-users] scheduler doesn't run ?

Per Jessen per at computer.org
Thu Jan 14 20:52:25 UTC 2016


Gary Buhrmaster wrote:

> On Thu, Jan 14, 2016 at 8:21 PM, Per Jessen <per at computer.org> wrote:
> .....
>> Both /tmp and /var/tmp are on real iron.
> 
> Mysql (in most configuration) will use /tmp for temporary table
> storage.  And if /tmp is marked with barriers, this can be slow.

mysql seems to be using /var/tmp in this setup. 
The physical setup of /var/tmp has not changed, ever.  

> I would recommend you either move /tmp to tmpfs (common
> in many more recent distros), or change the mysql temp table
> location to another tmpfs backed place, or specify enough
> memory in the mysql configuration such that the temp tables
> are never written to "disk" at all (or only in extreme situations).

I could try adding a separate disk on a separate io channel for tmp space, 
but it seems like a stab in the dark, when it was working fine up until 
Nov/Dec. 

>> This morning I did an optimize of "recordedseek" (millions of rows) which
>> seems to have improved things.  I'll report back later.
> 
> I could (very easily) be wrong, but I do not think the scheduler
> cares about recordedseek. 

I have no idea either, but scheduling has improved today, a lot. Gut 
feeling. Sofar, the longest was some 500seconds, yesterday a few went on for 
an hour. 

> On the other hand, the optimize
> script does optimize/analyze all the tables such that you
> may see improvements there.  Creating (more) accurate
> statistics for the DB engine to optimize its query can help a
> lot.

I'll try that too. 


/Per



More information about the mythtv-users mailing list