Re: [HPDD-discuss] time until qos_prio_free=0 has an impact
On Oct 8, 2014, at 10:46 AM, Mohr Jr, Richard Frank (Rick Mohr) wrote:
On Oct 8, 2014, at 1:26 AM, Michael Kluge <michael.kluge(a)tu-dresden.de>
wrote:
>> I thought the oss servers would dynamically create more threads up to
>> a limit of 512, so I would expect the load to max out around that
>> same point unless the number of threads has been increased on
>> purpose.
> Is it 512 per OST, I think. On the oss I have 12 active OSTs and about 3000 active
ll_ost_io threads.
Hmmm. That's interesting. I always assumed that the oss server threads were not tied
to any particular ost. Therefore, when the lustre manual explains how to increase the oss
thread count, I assumed that meant it was intended to increase the size of the total pool
of threads. Looking back at the manual, it doesn't seem to clearly explain one way or
the other.
Currently the threads are not tied to any particular OST or MDT service.
I think I remember there was a project in the past where for NUMA machines you were
supposed to be able to tie certain threads to certain OSTs, but there was nothing to
actually prevent the other threads from not hitting this pool should the original threads
gets all used up.
Bye,
Oleg