[gpfsug-discuss] SS 5.0.x and quota issues

Jaime Pinto pinto at scinet.utoronto.ca
Mon May 18 14:55:47 BST 2020


So Bob,
Yes, we too have observed an uncharacteristic lag on the correction of the internal quota accounting on GPFS since we updated from version 3.3 to version 4.x some 7-8 years ago. That lag remains through version 5.0.x as well. And it persisted through several appliances (DDN, G200, GSS, ESS and now DSS-G). In our university environment there is also a lot of data churning, in particular small files.

The workaround has always been to periodically run mmcheckquota on the top independent fileset to expedite that correction (I have a crontab script that measures the relative size of the in-dought columns on the mmrepquota report, size and inodes, and if either exceeds 2% for any USR/GRP/FILESET I run mmrepquota)

We have opened supports calls with IBM about this issue in the past, and we never got a solution, to possibly adjust some GPFS configuration parameter, and have this correction done automatically. We gave up.

And that begs the question: what do you mean by "... 5.0.4-4 ... that has a fix for mmcheckquota"? Isn't mmcheckquota zeroing the in-doubt columns when you run it?

The fix should be for gpfs (something buried in the code over many versions). As far as I can tell there has never been anything wrong with mmcheckquota.

Thanks
Jaime


On 5/18/2020 08:59:09, Cregan, Bob wrote:
> Hi,
>        At Imperial we  have been experiencing an issue with SS 5.0.x and quotas. The main symptom is a slow decay in the accuracy of reported quota usage when compared to the actual usage as reported by "du". This discrepancy can be as little as a few percent and as much as many  X100% . We also sometimes see bizarre effects such negative file number counts being reported.
> 
> We have been working with IBM  and have put in the latest 5.0.4-4 (that has a fix for mmcheckquota) that we have been pinning our hopes on, but this has not worked.
> 
> Is anyone else experiencing similar issues? We need to try and get an idea if this is an issue peculiar to our set up or a more general SS problem.
> 
> We are using user and group quotas in a fileset context.
> 
> Thanks
> 
> 
> *Bob Cregan*
> HPC Systems Analyst
> 
> Information & Communication Technologies
> 
> Imperial College London,
> South Kensington Campus London, SW7 2AZ
> T: 07712388129
> E: b.cregan at imperial.ac.uk
> 
> W: www.imperial.ac.uk/ict/rcs <http://www.imperial.ac.uk/ict/rcs>
> 
> _1505984389175_twitter.png @imperialRCS @imperialRSE
> 
> 1505983882959_Imperial-RCS.png
> 
> 
> 
> _______________________________________________
> gpfsug-discuss mailing list
> gpfsug-discuss at spectrumscale.org
> http://gpfsug.org/mailman/listinfo/gpfsug-discuss
> 

.
.
.        ************************************
           TELL US ABOUT YOUR SUCCESS STORIES
          http://www.scinethpc.ca/testimonials
          ************************************
---
Jaime Pinto - Storage Analyst
SciNet HPC Consortium - Compute/Calcul Canada
www.scinet.utoronto.ca - www.computecanada.ca
University of Toronto
661 University Ave. (MaRS), Suite 1140
Toronto, ON, M5G1M1
P: 416-978-2755
C: 416-505-1477



More information about the gpfsug-discuss mailing list