S
SteveInDelaware
A solid, stable platform for 2-1/2 years. Two weeks ago the system-managed pagefile grew to such a size that it filled the system drive and crashed the server for the first time.
I moved the pagefile to an unused 1T partition, leaving it system-managed.
This is a virtual machine running in Hyper-V with 96GB of RAM and 30 processors. Has run fine since we built it.
I am monitoring the pagefile, CPU, page faults, disk i/o, etc. with Task Manager, Resource Monitor, Performance Monitor and Process Explorer.
After rebooting, pagefile.sys is 256GB in size, with zero utilization. As time goes on, however, the usage of pagefile.sys goes up. And up. And up. And up until we get Event ID 2004 Resource Exhaustion Detector, low virtual memory condition. The server crashes shortly afterwards.
Typically (for years...) CPU utilization is ~40% and RAM utilization is about 45-50%. Performance is consistently good. No changes to CPU or RAM utilization is observed.
Logging users off and terminating tasks DOES NOT cause the pagefile usage to come back down. Whatever is getting swapped out is staying there.
Nothing except rebooting the server brings the pagefile usage back down.
I have not been able to find any tool or utility that will let me see what is actually taking up space in the pagefile. I now have to babysit the server, rebooting after-hours so that the pagefile is emptied.
1. Is there any tool to let me see what is going on inside the pagefile? The Microsoft Product Manager says no here --
Quote --
“There's no easy way to tell how much of a process has actually been paged out to the pagefile (i.e. actually stored in the pagefile thus using it). I am not aware of a tool that reports this information per process.”
2. OK, then ... how do you troubleshoot this problem?
3. Why is the pagefile not releasing unused pages?
Thank you for your kind assistance. What other helpful information can I provide?
Thanks again,
Steve in Delaware
Continue reading...
I moved the pagefile to an unused 1T partition, leaving it system-managed.
This is a virtual machine running in Hyper-V with 96GB of RAM and 30 processors. Has run fine since we built it.
I am monitoring the pagefile, CPU, page faults, disk i/o, etc. with Task Manager, Resource Monitor, Performance Monitor and Process Explorer.
After rebooting, pagefile.sys is 256GB in size, with zero utilization. As time goes on, however, the usage of pagefile.sys goes up. And up. And up. And up until we get Event ID 2004 Resource Exhaustion Detector, low virtual memory condition. The server crashes shortly afterwards.
Typically (for years...) CPU utilization is ~40% and RAM utilization is about 45-50%. Performance is consistently good. No changes to CPU or RAM utilization is observed.
Logging users off and terminating tasks DOES NOT cause the pagefile usage to come back down. Whatever is getting swapped out is staying there.
Nothing except rebooting the server brings the pagefile usage back down.
I have not been able to find any tool or utility that will let me see what is actually taking up space in the pagefile. I now have to babysit the server, rebooting after-hours so that the pagefile is emptied.
1. Is there any tool to let me see what is going on inside the pagefile? The Microsoft Product Manager says no here --
Quote --
“There's no easy way to tell how much of a process has actually been paged out to the pagefile (i.e. actually stored in the pagefile thus using it). I am not aware of a tool that reports this information per process.”
2. OK, then ... how do you troubleshoot this problem?
3. Why is the pagefile not releasing unused pages?
Thank you for your kind assistance. What other helpful information can I provide?
Thanks again,
Steve in Delaware
Continue reading...