On Thu, 7 Sep 2000, Eric Youngdale wrote:
> The oddness is this. We were observing stalls in the
> processing of commands that was traced to the fact that the
> queue had remained plugged for an excessive amount of time.
> The stalls last for about 5 seconds or so.
Which is the default sleep interval for kupdate, which
calls run_task_queue(&tq_disk) inside flush_old_buffers().
> Some investigation revealed that part of the answer is that
> the bdflush daemon essentially forces a bunch of dirty pages to
> be written to disk, but never bothers to unplug the queue when
> it is done.
Heh, you noticed this too? ;)
> My gut tells me that it is wrong for bdflush to not unplug
> the queue when it is done queueing I/O requests.
It is. Not unplugging the queue results in higher throughput
when running a benchmark load, but seems to really harm
system throughput (and cause stalls) in /real/ loads.
This is most likely due to the fact that in most real life
loads we have to write data and metadata all over the place
and write clustering isn't as effective as in benchmark
loads (where we write 500MB of "data" at full speed to one
file).
What about a "benchmark" value in /proc/sys/vm/bdflush to
remind the rest of the world of why we want to unplug the
device queue in bdflush() ? ;)
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Sep 07 2000 - 21:00:31 EST