Add subsystem metrics for the dispatcher #13989
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
SUMMARY
Connect #12776
Demo:
I started with a very long list of metrics that we could put in, but settled on these as providing unique value with each metric, and simple enough they can be understood. I feel like we added too much data with the task manager & callback receiver.
These have absolutely red hot implications for system stability. If the availability is sub-90%, then message processing will lag notably. Scale up events should also be infrequent, and they are not. This is already a reported finding, but this gives instrumentation for the perfscale team to quantify that and hold some regression coverage over it.
ISSUE TYPE
COMPONENT NAME
ADDITIONAL INFORMATION
I am trying to offer this as a model to use the metrics to make progress in performance. Here are patches that are intended to address these performance metrics:
dispatcher_availability
by a static amount by avoiding running the debug template periodicallydispatcher_pool_scale_up_events
(indeed, making it flat) and decreasingdispatcher_pool_worker_count
dispatcher_availability