Find peaks/troughs of memory usage during rollup #16224
Conversation
|
@miq-bot add_labels bug, providers/metrics |
|
@gtanzillo @Ladas please help review this. thanks |
|
Checked commit jameswnl@683c1e8 with ruby 2.3.3, rubocop 0.47.1, and haml-lint 0.20.0 |
| BURST_COLS.each do |col| | ||
| value = rt.send(col) | ||
| rollup_burst(col, new_perf[:min_max], rt.timestamp, value) | ||
| end |
There was a problem hiding this comment.
What did moving this section out of the loop above help with?
There was a problem hiding this comment.
It was embedded in the loop of Metric::Capture.capture_cols.each do |col| which wouldn't cover any derived* columns.
This logic is doing burst-rollup and should be targeting BURST_COL. It was ok before because BURST_COL was a subset of Metric::Capture.capture_cols, but now it is not.
| .maximum(:derived_memory_used) | ||
| end | ||
| perfs = VimPerformanceAnalysis.find_perf_for_time_period(self, "daily", :end_date => Time.now.utc, :days => Metric::LongTermAverages::AVG_DAYS) | ||
| perfs.collect(&:abs_max_derived_memory_used_value).compact.max |
There was a problem hiding this comment.
What about the old code was causing it to not pick the maximum value?
There was a problem hiding this comment.
The old code was getting the max of daily averages, not the realtime peaks/troughs which were lost during rollup. (The burst-rollup part change is now determining and storing these during rollup time)
|
Awesome thanks @jameswnl ! |
Part of the fix for https://bugzilla.redhat.com/show_bug.cgi?id=1468634.
This PR covers the memory usage.
(#16195 covers CPU)