Re: SBS 2003 Premium R2 server becomes more unstable, page file gr

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

Les,

Thanks again for your perseverance. Concerned that the server was going to
crash and wanting to make the most of a reboot (if circumstances required
it), I jumped on the server and did two things:

1. throttled WSUS SQL to 512MB
2. moved the PF to be system managed

The second option did not require a reboot, but expanded our PF size to
almost 12GB. The PF is currently sitting at about 9.45GB and seems to be
pretty happy right now. Would you suggest a reboot at the end of the
evening? I have a scheduled time to do a little server maintenance. I was
planning on reinstalling SP2 for exchange to ensure that it's on. Otherwise,
I'll let it sit, but if there is anything you'd like me to monitor or if
there are any queries you'd like me to run in ISA, let me know. Thanks again
for your help.

Seth

"Les Connor [SBS MVP]" wrote:

I was more concerned that you had file space A/V scanning things that should
be excluded. It appears not.

I think the immediate task is to get the page file down to a proper size.

Note that after a restart, it may take more than a day for your memory use
to stabilize, but it should stabilize and it's not uncommon for SBS boxes to
run for months without a restart, if it wasn't for those pesky patch
tuesdays.

--
Les Connor [SBS MVP]


"ipadl" <ipadl@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:B52C715B-2EEC-402D-8E0C-53F3EB6C3B48@xxxxxxxxxxxxxxxx
Les,

I've throttled SQL to 512kMB for each instance. . . actually, I'm not sure
if I've applied that to the WSUS instance but it hasn't grown above 120MB
at
this point. I haven't done anything to exchange, but there was some
evidence
in other literature that an unpatched exchange would leak memory at times
or
use it and not give any back to the pool for other apps. I need to check
the
SP version of Exchange.

As for anti-virus and anti-spam, not long after the server was up and
working (1-2 weeks), we implemented Antigen which scans Exchange, but this
may be all. We have found a few solutions that we'd like to implement,
but
haven't do to the server's instability. I'd ultimately like to have a
network managed AV/AS system. Is this one step that we should implement
right away?

Seth

"Les Connor [SBS MVP]" wrote:

What have you got for file space anti-virus scanning application, and
what
exclusions are set?

--
Les Connor [SBS MVP]


"ipadl" <ipadl@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:545AC080-883D-4C89-9857-3A6783DCFA30@xxxxxxxxxxxxxxxx
Les,

Thanks for the reply. I'll try to set the PF to be system managed this
evening and see what I find. Please read the last response to Dave, as
it
gives more info and insight into what happened after some patches,
hotfixes,
and downloads from last night. Thanks,

Seth

"Les Connor [SBS MVP]" wrote:

Switch to a system managed pagefile, as a suggestion.

--
Les Connor [SBS MVP]


"ipadl" <ipadl@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:A681F3C6-67DE-4541-838F-DF6C8AE583B9@xxxxxxxxxxxxxxxx
Dave,

I'm hoping that this is true and that as we've applied hotfixes that
we've
created a stable environment. At this point, we have the page file
set
up
as
follows:

4 GB RAM
Drive E (System Partition, Raid 1): 3 GB PF (min and max)
Drive F (File Partition, Raid 5): 3 GB PF (min and max)

The total page file size that we have set is 6GB. Under the PF
Useage
in
the Task Manager, we see the 9.29 GB (current value). This should
be
according to MS specs and your suggestion of the PF being 1.5x the
RAM
,
correct? The server hasn't crashed yet, but it has in the past.
It's
been
up currently for 2 days and 15 hours. Strangely enough (and
possible
unrelated), we've also heard on two occaisions the fans kick up to a
higher
rate when the server has gone down. This usually stays high for
several
minutes and then drops down to a lower rate again.

Thanks again for your help. I appreciate your suggestions. Where
should
we
go from here?

Seth

"Dave Nickason [SBS MVP]" wrote:

Pure speculation here - disregarding that VM Size column in Task
Manager,
the rest of your memory usage info seems consistent with a normally
functioning SBS. I just caught from your previous post that you
have
the
page file configured to 10 GB - is that right? My speculation
would
be
that
because the system has that size of page file available, it's just
going
ahead and using it. Kind of like a system that would be fine with
2
GB
of
RAM, but when you up it to 4 GB, it uses more for no other reason
than
that
it's there to be used.

I wonder what would happen if you set the page file to a fixed size
of
1.5
times RAM and rebooted. That would be a standard configuration,
and
my
guess is that the system would then just start using a more typical
amount
of page file.

I'd give that a try - configure a page file to have a fixed size
equal
to
1.5 times the amount of physical RAM in the server. Reboot to
apply
that
setting, and see what happens. If the system does not start
reporting
virtual memory or similar errors, I'd see if the instability is
remedied,
and if not, I'd troubleshoot it differently, looking for a
different
cause.


"ipadl" <ipadl@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:46D9D4B1-DB4E-4ABF-9F43-717BC2E403A8@xxxxxxxxxxxxxxxx
Dave,

Thanks for the reply. In the task manager, the Mem Useage
doesn't
account
for much. The VM Size is what seems to be getting large (Mem
Useage
vs.
VM
Size included below)! I didn't note the Commit Charge at the
time
that
I
had
taken that data and I deleted my screen shot after building the
spreadsheet,
but right now we're looking at a PF size of 9.34 GB with a Commit
Charge
of
9571M / 9749M.

I spoke to Microsoft again today about the hotfixes that you
suggested,
and
unlike before, they had a replacement for the hotfix from article
916089.
The new article is 927891. It didn't look like it was really all
that
similar and I won't apply it until this evening, but this is what
Microsoft
said was the fix.

Any other thoughts? The wuauserv is still part of the svchost
which
seems
to be rather large on the VM size. Could this still be the
culprit
to
this
memory leak? I'll see if the server crashes today. When the PF
gets
up
around 9.4 GB it gets really unstable.

Image Name Mem Usage (K) VM Size (K)
store.exe 609,584 623,068
svchost.exe 62,188 605,276
wspsrv.exe 27,120 596,512
w3wp.exe 34,508 591,664
lsass.exe 41,296 589,908
inetinfo.exe 33,392 584,112
mad.exe 21,856 570,804
winlogon.exe 3,904 567,600
services.exe 97,904 560,292
ntfrs.exe 1,852 560,028
sqlservr.exe 547,040 558,156
sqlagent.EXE 3,100 552,880
dfssvc.exe 4,744 552,792
crss.exe 3,096 550,936
tbsrvc.exe 2,764 550,496
sqlservr.exe 253,864 262,940
w3wp.exe 76,612 77,544
sqlservr.exe 39,820 50,524
mmc.exe 28,800 26,864
hpsmhd.exe 16,740 15,736
explorer.exe 20,052 13,664
wmiprvse.exe 14,932 13,052
exmgmt.exe 10,420 11,076
dns.exe 6,824 10,760
mspdmin.exe 13,836 10,252
explorer.exe 14,248 10,216
mssearch.exe 1,368 9,728
OWSTIMER.EXE 10,584 8,564
hpsmhd.exe 10,732 8,332
svchost.exe 8,456 8,272
ismserv.exe 7,096 7,364
wins.exe 6,192 7,340
wmiprvse.exe 9,884 6,676
winlogon.exe 2,600 5,864
svchost.exe 4,912 5,312
svchost.exe 6,444 4,788
spoolsv.exe 5,088 4,500
W3Prefch.exe 5,968 3,744
llssrv.exe 5,288 3,648
dmadmin.exe 4,512 3,544
svchost.exe 4,608 3,448
svchost.exe 3,864 3,376
msdtc.exe 4,268 2,208
sqlmangr.exe 5,136 2,096
sqlmangr.exe 5,220 2,084
smhstart.exe 3,120 2,040
crss.exe 4,560 1,924
isastg.exe 3,812 1,776
sbscrexe.exe 4,336 1,572
taskmgr.exe 4,356 1,560
tcpsvcs.exe 2,128 1,440
svchost.exe 3,596 1,424
sqlwriter.exe 3,604 1,268


"Dave Nickason [SBS MVP]" wrote:

So at the time you got the information below, the system was
showing 9
GB
PF
Usage? Those figures only add up to a little under 2 GB for
memory
usage,
and even peak is only a little over 2. What figures are you
showing
under
Commit Charge on the Performance tab?


"ipadl" <ipadl@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:B677EFD4-B9E5-4DB3-B280-729D1C338FAD@xxxxxxxxxxxxxxxx
Les et al,

We've applied those patches and it seems as though the page
file
is
continuing to do it's old thing. It is growing to
approximately
9
GB
(out
of
10GB that we have set aside) and sometimes growing larger,
choking
other
applications. SQL has been limited to 512KB. Are there any
other
ideas
as
to how to diagnose this problem? I'm going to try to insert a
spreadsheet
into this post with a bunch of Processes that my help us find
.

Relevant Pages

  • Re: Alocated Memory Error (SQLSERVR)
    ... but my understanding is that the MSDE instances are designed to ... and you don't want it grabbing all that RAM. ... running but non ever going about 200mb of memory usage. ... Server 2000 SP4, ...
    (microsoft.public.windows.server.sbs)
  • Re: Windows 2000 Server StandardEdition /PAE Question
    ... specifically the SQL Server 2000 Standard Edition on that server. ... His point is that applications can only use up to 2 GB or RAM on ... a process running under Windows 2000 or Windows Server 2003 can ... used) with some of the memory being physical memory and some being virtual ...
    (microsoft.public.win2000.general)
  • Re: Alocated Memory Error (SQLSERVR)
    ... FWIW I am not using sbs 2003 premium, ... non ever going about 200mb of memory usage. ... on what you should throttle the RAM to. ... Server 2000 SP4, ...
    (microsoft.public.windows.server.sbs)
  • Re: SBS 2003 - IIS Websites quit working
    ... thumb, not a law, and b) impossible if you have>GB RAM unless the ... NOT SBS). ... There's an algorithm in memory management that opportunistically uses ... Generally the server will go 3-5 days before the websites stop ...
    (microsoft.public.windows.server.sbs)
  • Re: Alocated Memory Error (SQLSERVR)
    ... what you should throttle the RAM to. ... Server 2000 SP4, ... The memory problem on my ... shared exchange server address book or anything exchange related). ...
    (microsoft.public.windows.server.sbs)