<div dir="ltr">If I look at the recent regression runs (<a href="https://build.gluster.org/job/centos7-regression/">https://build.gluster.org/job/centos7-regression/</a>), there is more than 50% failure in tests.<div><br></div><div>At least 90% of the failures are not due to the patch itself. Considering regression tests are very critical for our patches to get merged, and takes almost 6-7 hours now a days to complete, how can we make sure we are passing regression with 100% certainty ?</div><div><br></div><div>Again, out of this, there are only a few tests which keep failing, should we revisit the tests and see why it is failing? or Should we mark them as &#39;Good if it passes, but don&#39;t fail regression if the tests fail&#39; condition?</div><div><br></div><div>Some tests I have listed here from recent failures:</div><div><br></div><div><span style="white-space:pre-wrap">tests/bugs/core/multiplex-</span><span style="white-space:pre-wrap">limit-issue-151.t</span><br></div><div><span style="white-space:pre-wrap">tests/bugs/distribute/bug-</span><span style="white-space:pre-wrap">1122443.t +++</span><br></div><div><span style="white-space:pre-wrap">tests/bugs/distribute/bug-</span><span style="white-space:pre-wrap">1117851.t</span><span style="white-space:pre-wrap"><br></span></div><div><span style="white-space:pre-wrap">tests/bugs/glusterd/bug-</span><span style="white-space:pre-wrap">857330/normal.t +</span><span style="white-space:pre-wrap"><br></span></div><div><span style="white-space:pre-wrap">tests/basic/mount-nfs-auth.t +++++ </span><br></div><div><span style="white-space:pre-wrap">tests/basic/changelog/</span><span style="white-space:pre-wrap">changelog-snapshot.t</span><br></div><div><span style="white-space:pre-wrap">tests/basic/afr/split-brain-</span><span style="white-space:pre-wrap">favorite-child-policy.t</span><br></div><div><div><span style="white-space:pre-wrap">tests/basic/distribute/</span><span style="white-space:pre-wrap">rebal-all-nodes-migrate.t</span><span style="white-space:pre-wrap"><br></span></div><div></div></div><div><span style="white-space:pre-wrap">tests/bugs/glusterd/quorum-</span><span style="white-space:pre-wrap">value-check.t</span><br></div><div><span style="white-space:pre-wrap">tests/features/lock-</span><span style="white-space:pre-wrap">migration/lkmigration-set-</span><span style="white-space:pre-wrap">option.t</span><span style="white-space:pre-wrap"><br></span></div><div><span style="white-space:pre-wrap">tests/bugs/nfs/bug-1116503.t</span><span style="white-space:pre-wrap"><br></span></div><div><span style="white-space:pre-wrap">tests/basic/ec/ec-quorum-</span><span style="white-space:pre-wrap">count-partial-</span><span class="gmail-il" style="white-space:pre-wrap">failure</span><span style="white-space:pre-wrap">.t</span><span style="white-space:pre-wrap"><br></span></div><div><br></div><div>Considering these are just 12 of 750+ tests we run, Should we even consider marking them bad till they are fixed to be 100% consistent?</div><div><br></div><div>Any thoughts on how we should go ahead?</div><div><br></div><div>Regards,</div><div>Amar</div><div><br></div><div>(+) indicates a count, so moreĀ + you see against the file, more times that failed.</div><div><br></div></div>