potiuk commented on PR #37446:
URL: https://github.com/apache/airflow/pull/37446#issuecomment-1946967719

   Merged!
   
   That explains it indeed, It did not fail in your PR, because it would 
**generally** work without mocking (and worked for me as well locally) if run 
in isolation / on not so busy system. Generally PRs will only run a subset of 
tests that selective checks find relevant - so in the original PR only a small 
subsert of tests was run:
   
   ```
   Always
   Providers[amazon]
   
Providers[apache.beam,apache.cassandra,apache.flink,apache.spark,celery,cncf.kubernetes,common.sql,facebook,hashicorp,microsoft.azure,microsoft.mssql,mysql,openlineage,oracle,postgres,presto,sa
   Providers[google]
   ```
   
   In the new PR - I saw you added "full tests  needed" and the set of tests 
was the same as in `main`:
   
   ```
   Providers[-amazon,google]
   Core
   WWW
   CLI
   Other
   Serialization
   Always
   PythonVenv
   API
   BranchExternalPython
   BranchPythonVenv
   ExternalPython
   Operators
   PlainAsserts
   Providers[amazon]
   Providers[google]
   ```
   
   That's why original change was "green" because the set of tests run did not 
make the machine "busy enough" to trigger the flaky condition. In a way, our 
`main` (canary) tests are a good way to find such issues, because they run up 
to 8 parallel test suites at a time (out of the 16 above) in 8 complete 
docker-compose instances run on the same docker engine (each one with its own 
database). So we are REALLY stressing the poor AWS instance and that uncovers 
all kinds of races and flakiness (which as we saw here is a good idea).
   
   
   


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to