Red Hat Bugzilla – Bug 823666
Explicit entry and exit points to profile mcollective calls in the log.
Last modified: 2015-05-14 21:54:29 EDT
Description of problem:
The profile merge script (rhc-profiler-merge-report) relies on inferring from the logs when a call begins and ends. This is especially dicey for parallel calls which overlap.
Add specific log entries explicitly mark the start, end and duration of any call for profiling, parallel or not.
Version-Release number of selected component (if applicable):
The ss_* mcollective calls cannot be profiled. The parallel call data is inacurate.
Steps to Reproduce:
1. Do a benchmark run with profiling enabled
2. run rhc-profiler-merge-report on the resulting data
Missing or inaccurate call data.
Accurate call data.
I was able to work around this by having rhc-profiler-merge-report look for the log entry for YAML encoding the response if all else fails and add log entries to the mcollective log that it already knew to look for.
Passing to Q/A in case there's anything they'd like to verify; but this is basically just related to the benchmark Jenkins task.
For QE, we never tested profiling stuff. I remembered the user story about profiling stuff is marked with "no need QE". If I am right, can you help verify this bug and verify it?
No need for Q/E. Finishing the bug.