My completely wacky issue turned out to be a kernel bug! Thank you to those who offered advice. It only took Microsoft Support 28 days to find the answer (27 days to get me to collect increasingly unrelated information and escalate it through the usual levels, and then one day once it got escalated to someone that I could speak to as a peer), but it was a documented issue, so they did refund the support charge. It’s still unclear what my services or drivers are seemingly doing to provoke the issue.
The fix is covered by KB2619529 [https://support.microsoft.com/en-us/kb/2619529], and surprisingly is still not generally released.
Symptoms: Assume that you run an application or a service on a computer that is running Windows Vista or Windows Server 2008. The application or the service repeatedly commits and then releases lots of memory pages. In this situation, the memory manager may be unable to commit memory pages anymore.
Notes: When this issue occurs, many running applications or services may experience various symptoms.
To recover from this issue, you must restart the computer.
*** When this issue occurs, the CommittedBytes value in performance counter becomes a negative value. ***
Cause: This issue occurs because in some scenarios the memory manager incorrectly returns more memory pages than are actually allocated. This behavior causes the Commit Charge counter to underflow and become a very high value. Therefore, memory allocation fails and causes the issue that is mentioned in the “Symptoms” section.
Taed… Thanks A LOT for taking the time to close the loop on this.
Much appreciated.
Cool bug, too.
Peter
OSR
@OSRDrivers
Very interesting bug. Sadly, we may never know the real reason that caused the issue.
28 days is pretty good in my experience. You have to plow through a lot of
scripted info collection nonsense before the bug escalates to anyone who
can actually analyze anything.
Mark Roddy
On Wed, Jan 27, 2016 at 3:27 PM, Taed Wynnell wrote:
> My completely wacky issue turned out to be a kernel bug! Thank you to
> those who offered advice. It only took Microsoft Support 28 days to find
> the answer (27 days to get me to collect increasingly unrelated information
> and escalate it through the usual levels, and then one day once it got
> escalated to someone that I could speak to as a peer), but it was a
> documented issue, so they did refund the support charge. It’s still
> unclear what my services or drivers are seemingly doing to provoke the
> issue.
>
> The fix is covered by KB2619529 [
> https://support.microsoft.com/en-us/kb/2619529], and surprisingly is
> still not generally released.
>
> Symptoms: Assume that you run an application or a service on a computer
> that is running Windows Vista or Windows Server 2008. The application or
> the service repeatedly commits and then releases lots of memory pages. In
> this situation, the memory manager may be unable to commit memory pages
> anymore.
> Notes: When this issue occurs, many running applications or services may
> experience various symptoms.
> To recover from this issue, you must restart the computer.
> When this issue occurs, the CommittedBytes value in performance
> counter becomes a negative value.
> Cause: This issue occurs because in some scenarios the memory manager
> incorrectly returns more memory pages than are actually allocated. This
> behavior causes the Commit Charge counter to underflow and become a very
> high value. Therefore, memory allocation fails and causes the issue that is
> mentioned in the “Symptoms” section.
>
>
>
> —
> NTDEV is sponsored by OSR
>
> Visit the list online at: <
> http://www.osronline.com/showlists.cfm?list=ntdev>
>
> MONTHLY seminars on crash dump analysis, WDF, Windows internals and
> software drivers!
> Details at http:
>
> To unsubscribe, visit the List Server section of OSR Online at <
> http://www.osronline.com/page.cfm?name=ListServer>
></http:>