aboutsummaryrefslogtreecommitdiffstats
path: root/resources/tools/scripts/compare_perpatch.py
AgeCommit message (Collapse)AuthorFilesLines
2019-04-17Increase reliability of per patch perf jobVratko Polak1-26/+41
CSIT-1485 - Measurement results are inconsistent, creating false positives. - To avoid false positives, sensitivity has to be lowered. + Measurements are no more interleaved. + Measurement result are no longer filtered. + The job now runs faster. - Ip6base regressions smaller than 4% are tolerated. + Ip6base false positives can happen, but less than 10% of the time. - Part of job parameters is set on ci-management level. + Voter script output is improved. Change-Id: I0ec7d5df8b397daf7ff0277a1137ee9f36d8d866 Signed-off-by: Vratko Polak <vrpolak@cisco.com>
2019-02-20FIX: Static variables all over the place for per patchPeter Mikus1-5/+5
- Incredible why is it so hard to write scripts not hardcoding anything. Change-Id: I48cf2cc3d25f9665a9ff7ca26756d3b70cdd2f9c Signed-off-by: Peter Mikus <pmikus@cisco.com>
2018-10-22Per patch: multiple BMRR callsVratko Polak1-25/+56
This is to reduce sensitivity on testbed unstable performance. Also add topo_cleanupbefore every pybot run to avoid issues with VPP uninstall. TRACE prints are left there to simplify investigation of false positives if/when they happen. Change-Id: I9b0cdcfbbe4aa0735a0596746ac32c9e93af0897 Signed-off-by: Vratko Polak <vrpolak@cisco.com>
2018-08-30CSIT-1135: Scripts for VPP per-patch testingVratko Polak1-0/+85
+ Scripts do not rely on (other) bootstraps. + Perf verify bootstrap is also the new style sript now. + Scripts are divided to functions for better re-use. + Functions are sourced from small number of large "library" files. - Still using jumpavg from pipy. - Perpatch has specific simplified parsing (instead of the PAL one). - Bash style document is in a separate Change. Change-Id: If88fa528ce155ea86b614e3d77c0550b91bbdf11 Signed-off-by: Vratko Polak <vrpolak@cisco.com>