[SERVER-38068] Track successful command completion in FTDC Created: 09/Nov/18 Updated: 06/Dec/22 |
|
| Status: | Backlog |
| Project: | Core Server |
| Component/s: | Diagnostics |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Improvement | Priority: | Minor - P4 |
| Reporter: | Benjamin Caimano (Inactive) | Assignee: | Backlog - Service Architecture |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | sa-remove-fv-backlog-22 | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Issue Links: |
|
||||||||||||
| Assigned Teams: |
Service Arch
|
||||||||||||
| Participants: | |||||||||||||
| Description |
|
Currently, we track two buckets for command execution:
I believe there should be a third metric bucket:
|
| Comments |
| Comment by Bruce Lucas (Inactive) [ 15/Nov/18 ] |
|
total - failed counts commands that are in progress or have completed successfully. The "ok" metric would allow us to separate in progress from successfully completed. Also, for infrequently executed long-running commands (which in fact tend to be the troublemakers) the "total" metric and the "ok" metric will mark the start and the end of the command, which can be useful for diagnosis. |
| Comment by Danny Hatcher (Inactive) [ 15/Nov/18 ] |
|
Speaking to the "ok" metric, what is the benefit this provides? Isn't "ok" simply "total" - "failed? Are there situations in which a quick glance at those two existing metrics were not sufficient to diagnose an issue? |
| Comment by Kevin Pulo [ 15/Nov/18 ] |
|
Might also be good if we could also do this for opcounters (though perhaps not success/fail), eg. a "retired" metric for each opcounter, which gets incremented when each op ceases consuming resources (eg. returns results to the client, is killed, ...). |
| Comment by Bruce Lucas (Inactive) [ 12/Nov/18 ] |
|
I like this idea, but we should think about the impact on ftdc - it will add about 200 metrics by my count. Most of those will not change so delta will be 0 most of the time, so maybe ok in terms of data size, but it will have some impact on the downstream consumers of ftdc. Worth some thought. |
| Comment by Benjamin Caimano (Inactive) [ 09/Nov/18 ] |
|
I also believe that "total" is not a great name to contrast failed, but that may need to stick around for compatibility reasons. |