Resource exhaustion and issues with monitored memory metrics

Brass Contributor

Preface:  I do not know if this is the correct board for this

 

I'm having resource exhaustion issues but am unable to track down the cause.  Further, while monitoring my system during a recent instance of exhaustion, I'm seeing things that I don't understand.

 

System:

  • Physical RAM:  16 GB
  • Page file:  30 GB (static)
  • Total virtual memory:  46 GB

Perfmon stats during this event (and similar seen during previous events).  Screenshot below:  

 

  • Commit limit: 45.3 GB
  • % Committed Bytes In Use: 98.5%
  • Committed Bytes: 44.6 GB
  • Pool Nonpaged Bytes: 853.3 MB
  • Pool Paged Bytes: 1.7 GB
  • Page file % Usage: 3.7% (Peak: 7.3)

Additionally, RAM Available was 6+ GB (unfortunately forgot to track this in perfmon but noticed it in task manager.

 

MemIssue_PerfmonScreenshot.png

 

Side note: My understanding is % Commit Bytes in Use is the ratio of Committed Bytes to Commit Limit, with the limit being determined by the size of pagefile plus physical ram. In this case 46GB. The commit limit is 45.3 GB, per perfmon, so that seems accurate.

 

First Question: How can the % Committed Bytes In Use be so high, with such low pagefile and physical RAM usage?

 

Second Question: How can I be having such exhaustion issues with such low page file usage and so much available RAM? 

 

Third Question: Any thoughts as to what could be causing this exhaustion issue?

 

Additional notes/thoughts: While monitoring this over the last few months, my page file usage and available physical RAM remain consistent at or near the values noted above. This suggests to me I have plenty of resources to spare, always. The only thing I could think could be at play here is some sort of memory leak in the kernel space, but I haven't been able to see anything meaningful with poolmon with respect to paged/non-paged usage as those numbers are consistent with what is listed above as well.

 

Note:  while typing this post, the % Committed Bytes In Use spiked past 99.5%, while all other metrics  seemed to stay the same, and my entire system crashed/stopped responding.  This is what typically happens every time I have this exhaustion issue.

 

Update:  I installed the WinSDK so I could get debug symbols for Sysinternals Procexp to see what my paged limits were (don't know why it needs this for that but not for non-paged limits), but anyway, it's telling me my paged limit is 16TB!  Is that for real?  I don't see how that could even be possible.

 

Non-paged limit is only 16GB as I would expect is more likely given that's my RAM size.

0 Replies