[stgt] tgtd memory usage

FUJITA Tomonori fujita.tomonori at lab.ntt.co.jp
Mon Apr 4 16:25:06 CEST 2011


On Mon, 4 Apr 2011 14:44:40 +0300
Or Gerlitz <ogerlitz at mellanox.com> wrote:

> FUJITA Tomonori wrote:
> > On Wed, 30 Mar 2011 22:12:23 +0400
> > stepan.fedorov wrote:
> 
> >> Thanks all for attension, but question remains open: what can we do,
> >> to serve 1500-3000 LUNs from one node with no more then 24Gb of memory?
> 
> > Maybe the following patch works for you.
> 
> Yep, it looks something that has to do with the threads code consumes
> about 10MB per thread... Stepan, you should be able to see that using
> your script that allocates targets and measures the memory consumption
> if applying the following trivial patch, I hope that either Tomo or
> someone here will be able to address that soon
> 
> > diff --git a/usr/bs_thread.h b/usr/bs_thread.h
> > index d460032..3e3ba79 100644
> > --- a/usr/bs_thread.h
> > +++ b/usr/bs_thread.h
> > @@ -1,4 +1,4 @@
> > -#define NR_WORKER_THREADS      4
> > +#define NR_WORKER_THREADS      1
> > 
> >  typedef void (request_func_t) (struct scsi_cmd *);
> 
> Tomo, could you spare few words on the patch / approach you had in mind
> to address that? I see that your patch would only call  bs_thread_open
> once, e.g if (!init), where today it is called once per LUN... also I
> wasn't sure to follow why you've enlarged the number of threads from 4
> to 16, does this moves to global thread pool or alike?

Yeah, this patch uses 16 global threads instead of creating 4 threads
per lun. Needs some cleanups though.

We had better adjust the number of threads dynamically. I think that
the static approach is fine for now.


--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



More information about the stgt mailing list