[Alerting] fix flaky test for index threshold grouping #60792
Conversation
pmuellr
added
Feature:Alerting
v8.0.0
release_note:skip
| @@ -135,7 +135,8 @@ export default function alertTests({ getService }: FtrProviderContext) { | |||
| } | |||
|
|
|||
| // there should be 2 docs in group-0, rando split between others | |||
| expect(inGroup0).to.be(2); | |||
There was a problem hiding this comment.
In these tests, there should be two actions scheduled for every alert interval, one in one "group", the other in a different one, and we wait for 2 turns of the alert interval. There are always three groups of data in the index, but termSize will limit only two groups being returned - the "top" one, and one of the others (which we don't care about, as it's random-ish). For count (this test), group-0 has more docs, and so should be present in the output. For the sum (below), group-3 has larger values, and so should be present. For min (below sum), group-0 has smaller values, and so should be present.
While there should be 2 docs in the "top" group, the test failure this was written for got 3 at one point. Seems to be a very-rarely-flaky situation, so the check was changed from === 2 to > 0, to account for slop in the opposite direction as well (only getting 1 instead of 2).
|
started a flaky test runner: https://kibana-ci.elastic.co/job/kibana+flaky-test-suite-runner/278/ |
|
flaky test runner didn't find anything: https://kibana-ci.elastic.co/job/kibana+flaky-test-suite-runner/278/ , will merge in a bit ... |
resolves elastic#60744 This is a fairly complex test, with alerts that run actions that write to an index which we then do queries over. The tests didn't account for some slop in all that async activity, but now should be about as flake-free as they can be.
pmuellr
force-pushed
the
alerting/min-grouped-test-fail
branch
from
9ec6d51 to
0de2ffd
Compare
💚 Build SucceededHistory
To update your PR or re-run it, just comment with: |
resolves elastic#60744 This is a fairly complex test, with alerts that run actions that write to an index which we then do queries over. The tests didn't account for some slop in all that async activity, but now should be about as flake-free as they can be.
resolves #60744 This is a fairly complex test, with alerts that run actions that write to an index which we then do queries over. The tests didn't account for some slop in all that async activity, but now should be about as flake-free as they can be.
* master: (39 commits) [APM]Create custom link from Trace summary (elastic#59648) [ML] Fixing app clean up (elastic#60853) [SIEM] Use ECS categorisation for Authentication widgets (elastic#60734) [NP] Remove kbnUrl usage in discover/dashboard/visualize (elastic#60016) Skip failing test [Uptime]Update fetch effect failed action handling (elastic#60742) [npm] upgrade elastic/maki (elastic#60829) [Uptime] Add Settings Page (elastic#53550) [APM] service maps: avoid unnecesary `useDeepObjectIdentity` (elastic#60836) [Index management] Re-enable index template tests (elastic#60780) Fixed UI/UX issues: alerts delete confirmation, combobox behaviors (elastic#60703) [SIEM] Fix patching of ML Rules (elastic#60830) [APM] Service Map - Separate overlapping edges by rotating nodes (elastic#60477) [Alerting] fix flaky test for index threshold grouping (elastic#60792) [SIEM][Detection Engine] Adds test scripts for machine learning feature Flatten child api response for resolver (elastic#60810) Change "url" to "urls" in APM agent instructions (elastic#60790) [DOCS] Updates API requests and examples (elastic#60695) [SIEM] [Cases] Create case from timeline (elastic#60711) [Lens] Resetting a layer generates new suggestions (elastic#60674) ...
resolves #60744
This is a fairly complex test, with alerts that run actions that write to
an index which we then do queries over. The tests didn't account for some
slop in all that async activity, but now should be about as flake-free as they
can be.