Jef Spaleta

@djaglowski I think all the prom receiver flaky tests are symptoms of a race condition in how the commonly used testComponent func and the http server routine associated with mockPrometheus...

Okay I think I found it. There may be a wait group being used incorrectly in the Mock Prometheus server, with the Done call being done in a racy way....

Okay I can reproduce the problem locally quite easily with a one line change in the tests. Here's the reproducer change: I synthetically slow down the ServeHTTP function by introducing...

As promised, here a commented branch with the reproducer sleep in place and including additional diagnostic logging to try to understand what is going on with. https://github.com/jspaleta/opentelemetry-collector-contrib/tree/js/diagnose-flaky-prom-tests Someone who has...

Okay I refactor the new waitFor function to use the Server's pov about when it thinks its served enough requests to let the tests run. The diagnostic order of operations...

Okay good news. This looks like config that can be tuned. there is a config called scrape_interval that is too tight.. this needs to be increased for the test situation....

Okay I think I have this sorted out now. I'm going to prep a PR with the changes to scrape_interval and scrape_timeout so they allow for some slowness in the...

@djaglowski PR #12986 up with proposed fix

hey, So thinking of it like this: 1) Initially we can write it to target the sensu-client/sensu-agent http port so it will work with both 1.x and 2.x. 2) Then...

## Alternative backwards compatible implementation... Sort asset build by sha512 alphanumeric value when iterating over builds for matching build selection. This would provide deterministic build selection behavior for testing/troubleshooting without...