We should turn on report_interval and perhaps max_latency on most/all of the standard wtperf tests so we can monitor latency spikes. At a minimum, turning on max_latency so tests will fail if we hit a high latency will automatically allow Jenkins to flag it as a failure is a very fast change.
Work needed would be:
- Modify wtperf runner configs to turn those on.
- Add shell commands in Jenkins configs to extract info into properties files.
- Add plots to Jenkins configs to show latencies.
- Add plots to index.md for summary page.