On Tue, Aug 30, 2011 at 6:33 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Tue, Aug 30, 2011 at 4:37 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>>> If this is on Linux, I am surprised
>>>> that you didn't get killed by the lseek() contention problem on a
>>>> machine with that many cores.
>
>>> Hm ... now that you mention it, all of these tests have been using
>>> the latest-and-greatest unreleased RHEL kernels.
>
>> It should be pretty easy to figure it out, though. Just fire up
>> pgbench with lots of clients (say, 160) and run vmstat in another
>> window. If the machine reports 10% system time, it's fixed. If it
>> reports 90% system time, it's not.
>
> I ran it up to "pgbench -c 200 -j 200 -S -T 300 bench" and still see
> vmstat numbers around 50% user time, 12% system time, 38% idle.
> So no lseek problem here, boss. Kernel calls itself 2.6.32-192.el6.x86_64.
Eh, wait a minute. 38% idle time? Did you use a scale factor that
doesn't fit in shared_buffers? If so you're probably testing how fast
you pass BufFreelistLock around...
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company