3

I meet a very serious problem with the server. Our servers is very fast before. But this week, it becomes slower and slower. Everyday afterroon, the w3wp process would halt and we can not connect to the sharepoint again. When nobody use the sharepoint server, then we monitor the process, we found that the memory usage grows larger and larger. Then we check the sharepoint log, there would be memory leak. I want to know, how to idetify what cause the memory leak? And what cause the server become slower and slower?

Best Regards,

Improve this question
3
  • What is the memory usage of the w3wp process? After you restart the iis ( or the applicaton pool) is the site fast or slow?
    – Igal Serban
    Sep 12, 2009 at 19:36
  • 1
    If a server suddenly start degrading after a long period of working fine, then the first question you need to ask is "What Changed"
    – Zypher
    Jan 14, 2010 at 5:29
  • @Igal - Anonymous wanted to add "In Application server W3WP using 172MB 164MB Total RAM 4GB. In WFE server W3WP using 700MB 270MB 160MB. Total ram is 4GB"
    – Chris S
    Feb 6, 2012 at 14:17

6 Answers 6

Reset to default
3

have you written your own web parts, workflows or events?

If so, run the SPDisposeCheck tool to help you to check your assemblies that use the SharePoint API so that memory is handled correctly

code.msdn.microsoft.com/SPDisposeCheck

Or

Have you installed any updates this week check the control panel?

Improve this answer
1
  • SPDisposeCheck is not available anymore, all links are broken. Does anyone any update on this? Has it been moved or should we use something else now instead? I don't think I should start a new thread since the question is still the same, only the answers here went mostly invalid here, need a new one.
    – user14004
    Jun 30, 2023 at 7:33
1

Every afternoon at the same time? Look at what scheduled jobs you have around that time... perhaps disable all of them tomorrow and see if things improve. Somehow, you've got to narrow down which components are contributing to the problem.

Improve this answer
1

Also, check your task manager to see if a process, csrss.exe, is using high CPU >90%. There is a documented bug in MS where, if a user account gets corrupted, it will spin. The fix is kind of a 'non-fix' since it involves permanently deleting content that you may not be able to properly backup! Following is the link to Microsoft's MSDN site:

"csrss stands for 'client/server run-time subsystem.

This is the user-mode portion of the Win32 subsystem (with Win32.sys being the kernel-mode portion). Csrss is an essential subsystem that must be running at all times. Csrss is responsible for console windows, creating and/or deleting threads, and some parts of the 16-bit virtual MS-DOS environment."

Reference

http://support.microsoft.com/kb/263201

Improve this answer
0

Check app pool memory sizes for anything aberrant, and - in particular - the number of web gardens set. I'm willing to bet that you have something like 10 or so of the latter; if so reducing the number to 1 or 2 will instantly resolve things.

Improve this answer
0

Start by using performance counters to check memory used by the w3wp.exe and owstimer.exe processes. These processes are the only ones that should be related to SharePoint. (There is a list of many counters here but using all of them together probably won't help much.)

Once you have confirmed the problem is in one of these processes, verify that the problem is memory leaks within .NET code. The # Gen 2 Collections monitor in particular is a very likely sign that memory is not being deallocated. There's a good summary of other monitors here.

Finally, check for which custom code is leaking memory. As Chris mentioned, SPDisposeCheck is a very good way of checking whether memory is being disposed of correctly. You can run it against a particular custom DLL and find the culprit very quickly. Be aware that sometimes it gives false positives but never false negatives.

Improve this answer
0

Do you have a Firefox browser sitting idle on the server? I have found -- quite consistently -- that if I forget to close my Firefox browser in the evening, that by the next morning my server is crawling slow and if I let it go 24 hours, it is unresponsive.

(ed note: I use the browser for admin -- I don't need to surf with it, although I can and do without much impact.)

If I close my browser daily, it will run indefinitely.

The mouse won't even move and forget serving up web pages. My config is approximately this:

  • WinXP Home Edition with all the SPs
  • Apache 2.2
  • Tomcat 5.x.x
  • Java / JRE / JSDK at their respective latest.
  • Firefox automatic updates to the latest
  • Ancient "Intel Inside" and only 512Mb and 60Gb

I have looked for resource leaking (handles, memory, etc.) and memory is always the same ~75-80% level with moderate paging. i/o count is ridiculous and climbs 10K per refresh. Threads and handles are all about the same as when the problem is not occurring.

I have surfed the web for answers and bug reports, but nothing is jumping out at me.

Also, I don't know if any other browsers have the same effect.

Improve this answer

You must log in to answer this question.