[gpfsug-discuss] Steps for gracefully handling bandwidth reduction during network maintenance

Alex Chekholko alex at calicolabs.com
Mon Jun 17 17:51:27 BST 2019


Hi all,

My experience with MaxMBpS was in the other direction but it did make a
difference.  We had lots of spare network bandwith (that is, the network
was not the bottleneck) and in the course of various GPFS tuning it also
looked like the disks were not too busy, and the NSDs were not too busy, so
bumping up the MaxMBpS improved performance and allowed GPFS to do more.

Of course, this was many years ago on different GPFS version and hardware,
but I think it would work in the other direction.

It should also be very safe to try.

Regards,
Alex


On Mon, Jun 17, 2019 at 9:47 AM Christopher Black <cblack at nygenome.org>
wrote:

> The man page indicates that maxMBpS can be used to "artificially limit how
> much I/O one node can put on all of the disk servers", but it might not be
> the best choice. Man page also says maxmbps is in the class of mmchconfig
> changes take place immediately.
> We've only ever used QoS for throttling maint operations (restripes, etc)
> and are unfamiliar with how to best use it to throttle client load.
>
> Best,
> Chris
>
> On 6/17/19, 12:40 PM, "gpfsug-discuss-bounces at spectrumscale.org on
> behalf of Skylar Thompson" <gpfsug-discuss-bounces at spectrumscale.org on
> behalf of skylar2 at uw.edu> wrote:
>
>     IIRC, maxMBpS isn't really a limit, but more of a hint for how GPFS
> should
>     use its in-memory buffers for read prefetches and dirty writes.
>
>     On Mon, Jun 17, 2019 at 09:31:38AM -0700, Alex Chekholko wrote:
>     > Hi Chris,
>     >
>     > I think the next thing to double-check is when the maxMBpS change
> takes
>     > effect.  You may need to restart the nsds.  Otherwise I think your
> plan is
>     > sound.
>     >
>     > Regards,
>     > Alex
>     >
>     >
>     > On Mon, Jun 17, 2019 at 9:24 AM Christopher Black <
> cblack at nygenome.org>
>     > wrote:
>     >
>     > > Our network team sometimes needs to take down sections of our
> network for
>     > > maintenance. Our systems have dual paths thru pairs of switches,
> but often
>     > > the maintenance will take down one of the two paths leaving all
> our nsd
>     > > servers with half bandwidth.
>     > >
>     > > Some of our systems are transmitting at a higher rate than can be
> handled
>     > > by half network (2x40Gb hosts with tx of 50Gb+).
>     > >
>     > > What can we do to gracefully handle network maintenance reducing
> bandwidth
>     > > in half?
>     > >
>     > > Should we set maxMBpS for affected nodes to a lower value?
> (default on our
>     > > ess appears to be maxMBpS = 30000, would I reduce this to ~4000
> for 32Gbps?)
>     > >
>     > > Any other ideas or comments?
>     > >
>     > > Our hope is that metadata operations are not affected much and
> users just
>     > > see jobs and processes read or write at a slower rate.
>     > >
>     > >
>     > >
>     > > Best,
>     > >
>     > > Chris
>     > > ------------------------------
>     > > This message is for the recipient???s use only, and may contain
>     > > confidential, privileged or protected information. Any
> unauthorized use or
>     > > dissemination of this communication is prohibited. If you received
> this
>     > > message in error, please immediately notify the sender and destroy
> all
>     > > copies of this message. The recipient should check this email and
> any
>     > > attachments for the presence of viruses, as we accept no liability
> for any
>     > > damage caused by any virus transmitted by this email.
>     > > _______________________________________________
>     > > gpfsug-discuss mailing list
>     > > gpfsug-discuss at spectrumscale.org
>     > >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwICAg&c=C9X8xNkG_lwP_-eFHTGejw&r=DopWM-bvfskhBn2zeglfyyw5U2pumni6m_QzQFYFepU&m=2ioq3oT4gzOlIvyQRqkdZF0GWKv1APEBmstC48AyVdo&s=fvxPTdT1cVT7av_-vR5-3wVgjIzEpUP8OY8vGx0i5kc&e=
>     > >
>
>     > _______________________________________________
>     > gpfsug-discuss mailing list
>     > gpfsug-discuss at spectrumscale.org
>     >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwICAg&c=C9X8xNkG_lwP_-eFHTGejw&r=DopWM-bvfskhBn2zeglfyyw5U2pumni6m_QzQFYFepU&m=2ioq3oT4gzOlIvyQRqkdZF0GWKv1APEBmstC48AyVdo&s=fvxPTdT1cVT7av_-vR5-3wVgjIzEpUP8OY8vGx0i5kc&e=
>
>
>     --
>     -- Skylar Thompson (skylar2 at u.washington.edu)
>     -- Genome Sciences Department, System Administrator
>     -- Foege Building S046, (206)-685-7354
>     -- University of Washington School of Medicine
>     _______________________________________________
>     gpfsug-discuss mailing list
>     gpfsug-discuss at spectrumscale.org
>
> https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwICAg&c=C9X8xNkG_lwP_-eFHTGejw&r=DopWM-bvfskhBn2zeglfyyw5U2pumni6m_QzQFYFepU&m=2ioq3oT4gzOlIvyQRqkdZF0GWKv1APEBmstC48AyVdo&s=fvxPTdT1cVT7av_-vR5-3wVgjIzEpUP8OY8vGx0i5kc&e=
>
>
> ________________________________
>
> This message is for the recipient’s use only, and may contain
> confidential, privileged or protected information. Any unauthorized use or
> dissemination of this communication is prohibited. If you received this
> message in error, please immediately notify the sender and destroy all
> copies of this message. The recipient should check this email and any
> attachments for the presence of viruses, as we accept no liability for any
> damage caused by any virus transmitted by this email.
> _______________________________________________
> gpfsug-discuss mailing list
> gpfsug-discuss at spectrumscale.org
> http://gpfsug.org/mailman/listinfo/gpfsug-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20190617/139354d6/attachment-0002.htm>


More information about the gpfsug-discuss mailing list