How to identify runaway process / memory leak

[Guest]

I have an SBS2000 server that pretty much hangs after about 10 days of
running because it runs out of nonpaged pool space. It has been running fine
for about 5 years, and this started perhaps two months back. The first time,
I thought it was just an anomaly, but there has turned out to be a consistent
memory leak. Through PerfMon, I have identified the fact that both the
nonpaged bytes & the handle count steadily rise after a reboot but cannot
easily identify which process is responsible.

All the documentation I have seen seems to require me to have a suspect
process before I can determine whether it is, indeed, the problem. However,
how can I identify the runaway process if I have no idea which one is at
fault?

 

[Pegasus \(MVP\)]

I have an SBS2000 server that pretty much hangs after about 10 days of
running because it runs out of nonpaged pool space. It has been running
fine
for about 5 years, and this started perhaps two months back. The first
time,
I thought it was just an anomaly, but there has turned out to be a
consistent
memory leak. Through PerfMon, I have identified the fact that both the
nonpaged bytes & the handle count steadily rise after a reboot but cannot
easily identify which process is responsible.

All the documentation I have seen seems to require me to have a suspect
process before I can determine whether it is, indeed, the problem.
However,
how can I identify the runaway process if I have no idea which one is at
fault?

Watch the Task Manager. It will tell you which process consumes
ever increasing amounts of memory.

 

[Guest]

It creeps up so slowly (over days, not minutes) that it is difficult to get a
visual baseline on that. Is there any simple way to capture/export the task
manager data to a flat/delimited file so that I can compare side-by-side
after an 8-hour or 36-hour period?

 

[John John]

In Perfmon you can monitor the Non Paged Pool on the Processes
themselves. Select the processes that you want to monitor and then set
a log on them. Select to to log to a csv file and you will be able to
view the results in a spreadsheet, at a glance you will be able to see
which process is increasingly using the Non Paged Pool. The log
interval is set a 1 second by default, you should change that so that it
creates more manageable log files, the log could be set to log the
values every 5 minutes or so.

John