[SERVER-5799] Report only "hard" page faults on Windows Created: 09/May/12 Updated: 14/Jul/17 Resolved: 14/Jul/17 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Diagnostics |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Improvement | Priority: | Major - P3 |
| Reporter: | Tad Marshall | Assignee: | DO NOT USE - Backlog - Platform Team |
| Resolution: | Won't Fix | Votes: | 7 |
| Labels: | Windows, stats | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
Windows |
||
| Issue Links: |
|
||||||||||||
| Backwards Compatibility: | Fully Compatible | ||||||||||||
| Participants: | |||||||||||||
| Description |
|
On Windows, we are calling the Windows API GetProcessMemoryInfo() and returning the PageFaultCount as serverStatus.extra_info.page_faults. This value is the sum of "soft" and "hard" page faults, which makes the value not directly comparable with Linux and other OSs. We should instead report only "hard" page faults: page faults that cause a read from the page file or other file (such as a MongoDB database file). |
| Comments |
| Comment by Mira Carey [ 14/Jul/17 ] |
|
The work necessary to support this feature is more invasive than we'd like to attempt. Moving to wiredtiger has also rendered this less useful |
| Comment by Tad Marshall [ 22/Jun/12 ] |
|
Turning off backport flag until we see what the code changes look like. Code has not been written yet. |
| Comment by Tad Marshall [ 03/Jun/12 ] |
|
Adjusting issue type, estimate and "fix by version" based on Scott's observation above. If we need to collect this data from Event Tracing for Windows, it's a bigger project than I had imagined. Do-able, but non-trivial. |
| Comment by Eric Milkie [ 09/May/12 ] |
|
When I wrote this originally, I had looked into fetching just hard faults and it seemed hard so I gave up. System hard faults would be more interesting to have, perhaps in addition to process faults, so that we could compare. |
| Comment by Scott Hernandez (Inactive) [ 09/May/12 ] |
|
Not sure we can easily report this without using event tracing . Need to investigate on per process stats like this. It would be easy to get this for the system, which may be better than reporting all process faults (soft and hard). |