I have a problem on WS12 with a mirrored ReFS volume. Once I fill the
disk with IoMeter, I generate IO and after a short time, the system runs
out of memory. I have 24 GB of memory in the system and my disk driver
is using only 2.5 GB of np memory, so how do I find out who is using
memory at the application layer? With windbg, !poolused shows that only
2.9 GB of np memory is used with 1.2 GB of paged memory in use. That
means over 20 GB is consumed by app land. It is possibly IoMeter, but
the same scenario runs fine when NTFS is used instead of ReFS. !vm
shows iometer using on 6.7 MB of mem. I suspect there is something in
ReFS or Volume Manager which is doing this. Is there a debugger command
which will easily reveal who has horked all the memory?
((&->
use poolmon to take memory usage snapshots from time to time, say at 5 mins
intervals. once you get a few runs compare them to see which tag is
continously increasing over time, this is the most likely suspect.
Then using a text string finder, search for this tag value in the ReFS
binary on your system, if the tag is present in the binary this is the
driver causing it, else you will have to widen your search to all drivers
in system32 folder to find the culprit…
On Thu, Jan 24, 2013 at 10:52 AM, QuasiCodo wrote:
> I have a problem on WS12 with a mirrored ReFS volume. Once I fill the
> disk with IoMeter, I generate IO and after a short time, the system runs
> out of memory. I have 24 GB of memory in the system and my disk driver is
> using only 2.5 GB of np memory, so how do I find out who is using memory at
> the application layer? With windbg, !poolused shows that only 2.9 GB of np
> memory is used with 1.2 GB of paged memory in use. That means over 20 GB
> is consumed by app land. It is possibly IoMeter, but the same scenario
> runs fine when NTFS is used instead of ReFS. !vm shows iometer using on
> 6.7 MB of mem. I suspect there is something in ReFS or Volume Manager
> which is doing this. Is there a debugger command which will easily reveal
> who has horked all the memory?
>
> ((&->
>
> —
> NTDEV is sponsored by OSR
>
> OSR is HIRING!! See http://www.osr.com/careers
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.**cfm?name=ListServerhttp:
>
–
- ab</http:>
How about not reposting the same question with different subject line?
See subject “ReFS memory leak” posted on 1/24/13 2:01 PM (MST).
On 1/28/2013 9:57 AM, xxxxx@broadcom.com wrote:
How about not reposting the same question with different subject line?
That’s exactly what I did and the poolmon log files show that only 2.9
GB of non paged pool memory is in use at the time of the problem. Paged
pool is 1.2 GB. Since I have 24GB of physical memory, I figure this is
something in app land. So I used RAMMap every 5 minutes and under the
“Use Counts” tab, it shows “Mapped File” is 22GB. Under the “File
Summary” tab, E:\iobw.tst is using 22GB. The “Processes” tab shows that
there are no processes with more that 148MB, which belongs to
RAMMap64.exe itself. Iometer.exe is using 3MB and Dynamo.exe is using
3.3 MB.
Since the problem is “Mapped File”, can this be some sort of IoMeter
incompatibility under WS12? Strange it does not happen with NTFS.
((&->
On 1/28/2013 3:54 AM, A B wrote:
use poolmon to take memory usage snapshots from time to time, say at 5
mins intervals. once you get a few runs compare them to see which tag is
continously increasing over time, this is the most likely suspect.
Then using a text string finder, search for this tag value in the ReFS
binary on your system, if the tag is present in the binary this is the
driver causing it, else you will have to widen your search to all
drivers in system32 folder to find the culprit…
On Thu, Jan 24, 2013 at 10:52 AM, QuasiCodo > mailto:xxxxx> wrote:
>
> I have a problem on WS12 with a mirrored ReFS volume. Once I fill
> the disk with IoMeter, I generate IO and after a short time, the
> system runs out of memory. I have 24 GB of memory in the system and
> my disk driver is using only 2.5 GB of np memory, so how do I find
> out who is using memory at the application layer? With windbg,
> !poolused shows that only 2.9 GB of np memory is used with 1.2 GB of
> paged memory in use. That means over 20 GB is consumed by app land.
> It is possibly IoMeter, but the same scenario runs fine when NTFS
> is used instead of ReFS. !vm shows iometer using on 6.7 MB of mem.
> I suspect there is something in ReFS or Volume Manager which is
> doing this. Is there a debugger command which will easily reveal
> who has horked all the memory?
>
> ((&->
>
> —
> NTDEV is sponsored by OSR
>
> OSR is HIRING!! See http://www.osr.com/careers
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.__cfm?name=ListServer
> http:
>
>
>
>
> –
>
> - ab</http:></mailto:xxxxx>
It looks like memory is not being unlocked. What !vm says? Post the complete output.
On 1/28/2013 9:57 AM, xxxxx@broadcom.com wrote:
How about not reposting the same question with different subject line?
Missed the “not” – Doh!
On 1/28/2013 6:43 PM, xxxxx@broadcom.com wrote:
It looks like memory is not being unlocked. What !vm says? Post the complete output.
Here’s the output from !vm:
1: kd> !vm 20
*** Virtual Memory Usage ***
Physical Memory: 6289181 ( 25156724 Kb)
Page File: ??\C:\pagefile.sys
Current: 3407872 Kb Free Space: 3266204 Kb
Minimum: 3407872 Kb Maximum: 75497472 Kb
Type information missing error for MiSystemVaTypeCount
Available Pages: 26290 ( 105160 Kb)
ResAvail Pages: 5898073 ( 23592292 Kb)
Locked IO Pages: 0 ( 0 Kb)
Free System PTEs: 33556406 ( 134225624 Kb)
Modified Pages: 4318785 ( 17275140 Kb)
Modified PF Pages: 23 ( 92 Kb)
NonPagedPool Usage: 199916 ( 799664 Kb)
NonPagedPoolNx Usage: 30921 ( 123684 Kb)
NonPagedPool Max: 12220900 ( 48883600 Kb)
PagedPool 0 Usage: 166496 ( 665984 Kb)
PagedPool 1 Usage: 2374 ( 9496 Kb)
PagedPool 2 Usage: 616 ( 2464 Kb)
PagedPool 3 Usage: 558 ( 2232 Kb)
PagedPool 4 Usage: 612 ( 2448 Kb)
PagedPool Usage: 170656 ( 682624 Kb)
PagedPool Maximum: 100663296 ( 402653184 Kb)
Session Commit: 7250 ( 29000 Kb)
Shared Commit: 21620 ( 86480 Kb)
Special Pool: 0 ( 0 Kb)
Shared Process: 4825 ( 19300 Kb)
PagedPool Commit: 170719 ( 682876 Kb)
Driver Commit: 47244647209 (188978588836 Kb)
Committed pages: 680522 ( 2722088 Kb)
Commit limit: 7141149 ( 28564596 Kb)
Total Private: 101174 ( 404696 Kb)
0a3c ServerManager. 32440 ( 129760 Kb)
053c dwm.exe 9604 ( 38416 Kb)
094c explorer.exe 8710 ( 34840 Kb)
03a4 svchost.exe 5558 ( 22232 Kb)
0184 svchost.exe 3859 ( 15436 Kb)
0664 svchost.exe 3705 ( 14820 Kb)
0660 perfmon.exe 3634 ( 14536 Kb)
0518 svchost.exe 3359 ( 13436 Kb)
0644 Taskmgr.exe 3053 ( 12212 Kb)
0358 svchost.exe 2895 ( 11580 Kb)
0b24 notepad.exe 1856 ( 7424 Kb)
0bd4 mmc.exe 1757 ( 7028 Kb)
082c taskhostex.exe 1755 ( 7020 Kb)
043c svchost.exe 1735 ( 6940 Kb)
0ab0 Iometer.exe 1697 ( 6788 Kb)
03e0 svchost.exe 1627 ( 6508 Kb)
09bc WmiPrvSE.exe 1590 ( 6360 Kb)
0264 lsass.exe 1426 ( 5704 Kb)
02ac conhost.exe 1106 ( 4424 Kb)
025c services.exe 1076 ( 4304 Kb)
02d4 svchost.exe 801 ( 3204 Kb)
0c1c WerFault.exe 773 ( 3092 Kb)
0308 svchost.exe 767 ( 3068 Kb)
04e0 spoolsv.exe 760 ( 3040 Kb)
08ac WerFault.exe 746 ( 2984 Kb)
0974 WerFault.exe 708 ( 2832 Kb)
0b34 msdtc.exe 681 ( 2724 Kb)
01bc WmiPrvSE.exe 578 ( 2312 Kb)
03a0 csrss.exe 515 ( 2060 Kb)
0750 winlogon.exe 370 ( 1480 Kb)
0dd0 cmd.exe 332 ( 1328 Kb)
01a8 csrss.exe 324 ( 1296 Kb)
01f4 wininit.exe 261 ( 1044 Kb)
072c svchost.exe 201 ( 804 Kb)
0a68 taskhost.exe 193 ( 772 Kb)
0f64 poolmon.exe 128 ( 512 Kb)
0e64 conhost.exe 122 ( 488 Kb)
05a0 taskhostex.exe 122 ( 488 Kb)
0028 taskhost.exe 122 ( 488 Kb)
0310 csrss.exe 84 ( 336 Kb)
013c smss.exe 76 ( 304 Kb)
0a60 smss.exe 39 ( 156 Kb)
0004 System 29 ( 116 Kb)
0720 TSTheme.exe 0 ( 0 Kb)
06a0 wlrmdr.exe 0 ( 0 Kb)
045c cmd.exe 0 ( 0 Kb)
0430 smss.exe 0 ( 0 Kb)
032c LogonUI.exe 0 ( 0 Kb)
WHy did you have to limit pagefile size?