search

scylladb / python-driver

ScyllaDB Python Driver, originally DataStax Python Driver for Apache Cassandra
https://python-driver.docs.scylladb.com
Apache License 2.0
70 stars 42 forks source link

Enable tests/integration/standard #219

Closed Lorak-mmk closed 1 year ago

Lorak-mmk commented 1 year ago

Some integration tests were disabled in CI, and so have rotten a bit and were not passing - most notably, our shard-awareness tests. This PR fixes / disables failing tests in tests/integration/standard and enables them in CI.

There are also some minor general fixes in tests (strict xfail, pip install -e). See commit messages for descriptions of each change.

Lorak-mmk commented 1 year ago

Rebased on master, reverted one changed requested by @fruch. There was one more failing test, and it passes for me locally - if it fails again I'll investigate.

Regarding removal of "querying locally" assert: I propose to merge this without restoring the assert - as it causes the test to fail. I'll open an issue about it, and look into it after coming back from PTO.

Lorak-mmk commented 1 year ago

I don't get why this failed - the branch contains the --smp 1 fix that was merged to master.

@edit 2023-03-30 13:26:42.165 DEBUG [thread:57]: Connected to 2/2 shards on host 127.0.0.2:9042 (0 missing or needs replacement) <- a message from logs. There should be 1 shard, maybe the fix doesn't work for some reason?

Lorak-mmk commented 1 year ago

Oh, it's probably not creating a new cluster but reusing the existing one

Lorak-mmk commented 1 year ago

Ok, so test_authentication_misconfiguration.py seems to be flaky, I'll have to fix it...

Lorak-mmk commented 1 year ago

Rebased on master - to see if my (now reverted) PR wasn't the cause of test failures.

Lorak-mmk commented 1 year ago

I don't want to merge it yet. There are flaky tests which I want to disable (and open issues about them), but the old logs expired so I rerun CI a few times - and now they started to pass of course ;_;

Lorak-mmk commented 1 year ago

I don't get it - there were 1-3 tests that were failing with high probability, but now there are I think 5 successful runs in a row. Magic? I guess we can merge this, if failures reappear I'll disable the tests and open the issues.