Detecting leaks...how?

  • Thread starter Thread starter Fran
  • Start date Start date
F

Fran

I have a Windows 2000 server that has been having troubles now at
least twice a week where it needs to be rebooted. I get errors 2020
and 2016 (The server was unable to allocate virtual memory. )

I suspect some app has a memory leak and the server is running out of
RAM and other resources but I don't know how to pinpoint what app (or
service) is causing this to happen.

Are there any tools or utilities I can run that will help me figure
this out?

Fran
 
In
Fran said:
I have a Windows 2000 server that has been having troubles now at
least twice a week where it needs to be rebooted. I get errors 2020
and 2016 (The server was unable to allocate virtual memory. )

I suspect some app has a memory leak and the server is running out of
RAM and other resources but I don't know how to pinpoint what app (or
service) is causing this to happen.

Are there any tools or utilities I can run that will help me figure
this out?

Fran

One simple way would be to watch your task manager and see what teh memory
usage is. If you can't tell, try to start shutting down services until you
see which one releases it.

Check your Event viewer for errors logged.

Setup a Performance monitor log and add the memory object and the associated
counters. Also add your services installed. If they are MS apps + services,
you will see specific counters, if not, you will just see the executables.
Let it log for a few days or longer and then check out the results.

If there are any rogue apps running (maybe your machine got 'pubbed' with an
ftp server on it?) that will drain it. Also depends on what is running on
the machine (Exchange, SQL, is it a DC, etc).

So there is no simple answer here. Need to do some investigating.

Maybe these may help as well:

194655 - Using the PHD Class to Isolate Memory Leaks:
http://support.microsoft.com/default.aspx?scid=kb;en-us;q194655

Monitoring your IIS Web servers - Builder UK:
http://uk.builder.com/architecture/web/0,39026570,20265586,00.htm


--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2003 & 2000, MCSA 2003 & 2000, MCSE+I, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.
 
If you're lucky, it's Norton Anti Virus - see:

We're using CA e-TrustIT so it's not Norton's fault :(
If you're less lucky, some process visible in Task Manager will obviously be
using too much memory. If so, you may be able to configure a temporary fix
by scheduling a periodic stop/start of the associated service.

Short of watching constantly there's nothing that really jumps out at
you in Task Manager. The only service that's really pigging memory is
the database server (it uses 238M of memory but not all of it is RAM.
Most of it (better than 90%) is VM) The other services that are eating
memory but in a much less fashion are:
Backup Exec (about 60 meg when you add up all the subordinate things
running along with SQL MSDE.)
e-TrustIT takes up about 10 meg

So... looks like I"m really unlucky. I'll try the Poolmon.exe program
(is this Jamican?)


Thanks, Doug, for the tips. I hope we can track this down quickly. I
did activate some monitors but I don't know that they're going to tell
me what the process is that went south.

Fran
 
Stab in the dark - if this server is hosting a web based front end for the
database, try:

iisreset /stop
iisreset /start

Good luck, mon

Doug Sherman
MCSE Win2k/NT4.0, MCSA, MCP+I, MVP
 
By now you must have found Task Manager makes this pretty
easy.

Also note that when Process tab is selected (this is the one
you primarily use for such problems) then menu:

View\Select Columns

....lets you add a variety of columns that are not displayed
by default, including things like NonPaged pool memory
which is much more critical than other memory most of
the time....

It's probably obvious but those columns are SORT BUTTONS
too, so you can sort by ascending, click again descending,
usage on any pertinent column.
 
To all of you, thanks! I have started some counters to log what's
going on. I'm running this for a week (although lately I'll be lucky
to get this server to run a week without locking up.)

I tried the task manager but after watching it for 30 minutes I didn't
see anything that jumped out. The only "big" app that was running was
the database server (Pervasive) and that went up and down in usage. It
uses a combination of Virtual Memory and RAM and I really didn't see
much of a fluctiation in all that.

I really DO appreciate everyone's assistance and ideas on this. WHEN I
find the leak(s) or whatever it is that's causing this server to lose
page pool memory (Event ID 2020) I'll post the results (hopefully to
help someone else!)

Fran
 
To all of you, thanks! I have started some counters to log what's
going on. I'm running this for a week (although lately I'll be lucky
to get this server to run a week without locking up.)

That's a better method if you have a SLOW leak.
I tried the task manager but after watching it for 30 minutes I didn't
see anything that jumped out. The only "big" app that was running was
the database server (Pervasive) and that went up and down in usage. It
uses a combination of Virtual Memory and RAM and I really didn't see
much of a fluctiation in all that.

Task manager is more for an "instaneous" look -- it's strength
is that you don't need to "setup" the counters.

You might consider "alerts" on some particular counter(s) --
not all counters make sense for alerts so you might need to think
about which ones and when to alert carefully, but this might be
good if you don't know "when" the problem will occur but
expect it will occur more often than you will be certain to
check.

I call counters useful for alerts, "threshold counters" -- those that
reach some value and generally stay above (or below) it. Those
that bounc up and down (like CPU) are pretty worthless for alerts.)
I really DO appreciate everyone's assistance and ideas on this. WHEN I
find the leak(s) or whatever it is that's causing this server to lose
page pool memory (Event ID 2020) I'll post the results (hopefully to
help someone else!)

Perhaps an alert on "free paged pool memory". (Or whatever the
precise count is called.

You do realize that you can monitor the server using the SysMon-
perfmon on another machine, right?
 
Hi,

we have the same Problem on 3 W2k Servers with Service Pack 3/4.
etrust is running on this machines too.
It seems to be, that etrust is a part of this Problem but not th
cause.

On our machines the ID 2020 coms with a full Backup over the Network
If etrust is running the Problem becomes earlier.

You can also resolve the Problem if you copy a large number off File
from th Server to an network Share with xcopy or Lxcopy.

maybe it helps to yo


-
leak
 
Back
Top