potiuk commented on PR #35185: URL: https://github.com/apache/airflow/pull/35185#issuecomment-1806765381
And BTW. My wild guess - without looking at the changes/code is that this is some kind of race condition - where the service account is either not created yet or maybe while being deleting and recreated when it is needed. The tests we run in CI are (quite a bit deliberately) running in a rather constrained environment - we run up to 3 parallel Kind clusters and 3 parallel clusters at the same time - this means that we are stretching the limits of the machine it runs on - it has 8 cores and 64 GB memory, but it still has a lot of I/O contention and it might simply run a lot slower at times when a lot of other things happen - and that's not only good to squeeze everything out of that machine (and get the test finish as soon as possible) but also this is ideal environment where those kinds of race conditions happen - this is why our flaky tests are more flaky in CI usually, because the resources for single run are limited. This way we can also detect REAL issues - that we would only find out when our users try to run Airflow on a "minimum hardware" they pay as little as possible for. And this error smells like one of those cases where if you try to install helm in such limited resource environment (with these namespace modifications) it might fall into one of those race conditions. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@airflow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
