Hi,
One thing that springs to mind is maybe the DPM requests database
cleanup was running. Perhaps it had been configured but not activated
until the upgrade/yaim config triggered it. It seems to match the heavy
database load out of nowhere that then just stops.
John
On 21/01/2013 12:15, John Hill wrote:
> Hi Stephen, Sam,
> Thanks for the suggestions. I did increase innodb_buffer_pool_size
> during Friday to 2GB and it seemed to have no effect. However at about
> 09:00 on Saturday the load quickly dropped to a very low value and has
> been like that since (except for a couple of brief excursions to a load
> of 1). Either some external operation which I've not seen before really
> did start at the same time as the upgrade, or the upgrade caused a
> strange knock-on activity which look 36 hours to complete. I haven't yet
> had chance to see if I can find a correlation with external activities.
> Thanks for all the helpful suggestions from various people - much
> appreciated.
>
> John
>
> On 21/01/2013 11:37, Sam Skipsey wrote:
>> Indeed, the DPM tuning advice (written mostly by Alessandra and I)
>> recommends having a big buffer pool size for precisely that reason.
>> Try setting it to at least 1GB, but up to 4GB if you need to.
>>
>> Sam
>>
>>
>> On 21 January 2013 11:32, Stephen Jones <[log in to unmask]
>> <mailto:[log in to unmask]>> wrote:
>>
>> On 01/18/2013 12:15 PM, John Hill wrote:
>>
>> About 14GB. 8GB memory was fine with DPM 1.8.3, so if it is
>> inadequate
>> now then something significant has changed - and there's
>> nothing in
>> the Release Notes highlighting the need for a larger machine for
>> 1.8.4 or 1.8.5.
>>
>>
>> Hi John,
>>
>> I don't know very much about mysql, but Oracle would struggle
>> if the "buffer cache" was too small. It would have to keep paging
>> blocks out to fit new ones in, which made it waity. Maybe check that
>> you have decent buffers on your DB. I know there are some
>> settings in the /etc/my.cfg e.g. maybe
>> set-variable=innodb_buffer___pool_size=8000M,
>> (or query_cache_size=16000000???)
>>
>> If you have 8GB of mem, some DBA's recommend using half of it
>> for the buffer cache (assuming nothing else but the OS runs there).
>>
>> That's very naive, so I guess you'd have to select an appropriate
>> size
>> if other services run there. I dunno - just a few ideas...
>>
>> Steve
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> --
>> Steve Jones [log in to unmask] <mailto:[log in to unmask]>
>> System Administrator office: 220
>> High Energy Physics Division tel (int): 42334
>> Oliver Lodge Laboratory tel (ext): +44 (0)151 794
>> 2334 <tel:%2B44%20%280%29151%20794%202334>
>> University of Liverpool http://www.liv.ac.uk/physics/__hep/
>> <http://www.liv.ac.uk/physics/hep/>
>>
>>
--
John Bland [log in to unmask]
Research Fellow office: 220
High Energy Physics Division tel (int): 42911
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 2911
University of Liverpool http://www.liv.ac.uk/physics/hep/
"I canna change the laws of physics, Captain!"
|