<font size=3 face="Arial">AFAIK you can increase the pagepool size dynamically
but you cannot shrink it dynamically. To shrink it you must restart
the GPFS daemon. Also, could you please provide the actual pmap
commands you executed?</font><br><br><font size=2 face="sans-serif">Regards, The Spectrum Scale (GPFS) team<br><br>------------------------------------------------------------------------------------------------------------------<br>If you feel that your question can benefit other users of Spectrum
Scale (GPFS), then please post it to the public IBM developerWroks Forum
at </font><a href="https://www.ibm.com/developerworks/community/forums/html/forum?id=11111111-0000-0000-0000-000000000479"><font size=2 face="sans-serif">https://www.ibm.com/developerworks/community/forums/html/forum?id=11111111-0000-0000-0000-000000000479</font></a><font size=2 face="sans-serif">.
<br><br>If your query concerns a potential software error in Spectrum Scale (GPFS)
and you have an IBM software maintenance contract please contact 1-800-237-5511
in the United States or your local IBM Service Center in other countries.
<br><br>The forum is informally monitored as time permits and should not be used
for priority messages to the Spectrum Scale (GPFS) team.</font><br><br><br><br><font size=1 color=#5f5f5f face="sans-serif">From:
</font><font size=1 face="sans-serif">Aaron Knister <aaron.s.knister@nasa.gov></font><br><font size=1 color=#5f5f5f face="sans-serif">To:
</font><font size=1 face="sans-serif"><gpfsug-discuss@spectrumscale.org></font><br><font size=1 color=#5f5f5f face="sans-serif">Date:
</font><font size=1 face="sans-serif">02/22/2018 10:30 PM</font><br><font size=1 color=#5f5f5f face="sans-serif">Subject:
</font><font size=1 face="sans-serif">Re: [gpfsug-discuss]
pagepool shrink doesn't release all memory</font><br><font size=1 color=#5f5f5f face="sans-serif">Sent by:
</font><font size=1 face="sans-serif">gpfsug-discuss-bounces@spectrumscale.org</font><br><hr noshade><br><br><br><tt><font size=2>This is also interesting (although I don't know what
it really means). <br>Looking at pmap run against mmfsd I can see what happens after each step:<br><br># baseline<br>00007fffe4639000 59164K 0K 0K
0K 0K ---p [anon]<br>00007fffd837e000 61960K 0K 0K
0K 0K ---p [anon]<br>0000020000000000 1048576K 1048576K 1048576K 1048576K 0K
rwxp [anon]<br>Total: 1613580K 1191020K 1189650K 1171836K
0K<br><br># tschpool 64G<br>00007fffe4639000 59164K 0K 0K
0K 0K ---p [anon]<br>00007fffd837e000 61960K 0K 0K
0K 0K ---p [anon]<br>0000020000000000 67108864K 67108864K 67108864K 67108864K
0K rwxp [anon]<br>Total: 67706636K 67284108K 67282625K
67264920K 0K<br><br># tschpool 1G<br>00007fffe4639000 59164K 0K 0K
0K 0K ---p [anon]<br>00007fffd837e000 61960K 0K 0K
0K 0K ---p [anon]<br>0000020001400000 139264K 139264K 139264K 139264K 0K
rwxp [anon]<br>0000020fc9400000 897024K 897024K 897024K 897024K 0K
rwxp [anon]<br>0000020009c00000 66052096K 0K 0K
0K 0K rwxp [anon]<br>Total: 67706636K 1223820K 1222451K 1204632K
0K<br><br>Even though mmfsd has that 64G chunk allocated there's none of it <br>*used*. I wonder why Linux seems to be accounting it as allocated.<br><br>-Aaron<br><br>On 2/22/18 10:17 PM, Aaron Knister wrote:<br>> I've been exploring the idea for a while of writing a SLURM SPANK
plugin <br>> to allow users to dynamically change the pagepool size on a node.
Every <br>> now and then we have some users who would benefit significantly from
a <br>> much larger pagepool on compute nodes but by default keep it on the
<br>> smaller side to make as much physmem available as possible to batch
work.<br>> <br>> In testing, though, it seems as though reducing the pagepool doesn't
<br>> quite release all of the memory. I don't really understand it because
<br>> I've never before seen memory that was previously resident become
<br>> un-resident but still maintain the virtual memory allocation.<br>> <br>> Here's what I mean. Let's take a node with 128G and a 1G pagepool.<br>> <br>> If I do the following to simulate what might happen as various jobs
<br>> tweak the pagepool:<br>> <br>> - tschpool 64G<br>> - tschpool 1G<br>> - tschpool 32G<br>> - tschpool 1G<br>> - tschpool 32G<br>> <br>> I end up with this:<br>> <br>> mmfsd thinks there's 32G resident but 64G virt<br>> # ps -o vsz,rss,comm -p 24397<br>> VSZ RSS COMMAND<br>> 67589400 33723236 mmfsd<br>> <br>> however, linux thinks there's ~100G used<br>> <br>> # free -g<br>>
total used
free shared buffers
cached<br>> Mem: 125
100 25
0 0
0<br>> -/+ buffers/cache:
98 26<br>> Swap:
7 0
7<br>> <br>> I can jump back and forth between 1G and 32G *after* allocating 64G
<br>> pagepool and the overall amount of memory in use doesn't balloon but
I <br>> can't seem to shed that original 64G.<br>> <br>> I don't understand what's going on... :) Any ideas? This is with Scale
<br>> 4.2.3.6.<br>> <br>> -Aaron<br>> <br><br>-- <br>Aaron Knister<br>NASA Center for Climate Simulation (Code 606.2)<br>Goddard Space Flight Center<br>(301) 286-2776<br>_______________________________________________<br>gpfsug-discuss mailing list<br>gpfsug-discuss at spectrumscale.org<br></font></tt><a href="https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwIGaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=IbxtjdkPAM2Sbon4Lbbi4w&m=OrZQeEmI6chBdguG-h4YPHsxXZ4gTU3CtIuN4e3ijdY&s=hvVIRG5kB1zom2Iql2_TOagchsgl99juKiZfJt5S1tM&e="><tt><font size=2>https://urldefense.proofpoint.com/v2/url?u=http-3A__gpfsug.org_mailman_listinfo_gpfsug-2Ddiscuss&d=DwIGaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=IbxtjdkPAM2Sbon4Lbbi4w&m=OrZQeEmI6chBdguG-h4YPHsxXZ4gTU3CtIuN4e3ijdY&s=hvVIRG5kB1zom2Iql2_TOagchsgl99juKiZfJt5S1tM&e=</font></tt></a><tt><font size=2><br><br></font></tt><br><br><BR>