Monitor pods by labels instead of names

[dimberman]

Make sure you have checked all steps below.

Jira

• My PR addresses the following Airflow Jira issues and references them in the PR title. For example, "[AIRFLOW-XXX] My Airflow PR"
  • https://issues.apache.org/jira/browse/AIRFLOW-5589
  • In case you are fixing a typo in the documentation you can prepend your commit with [AIRFLOW-XXX], code changes always need a Jira issue.
  • In case you are proposing a fundamental code change, you need to create an Airflow Improvement Proposal (AIP).
  • In case you are adding a dependency, check if the license complies with the ASF 3rd Party License Policy.

Description

• Here are some details about my PR, including screenshots of any UI changes:

To prevent situations where the scheduler starts a
second k8sPodOperator pod after a restart, we now check
for existing pods using kubernetes labels

Tests

• My PR adds the following unit tests OR does not need testing for this extremely good reason:

Commits

• My commits all reference Jira issues in their subject lines, and I have squashed multiple commits if they address the same issue. In addition, my commits follow the guidelines from "How to write a good git commit message":
  1. Subject is separated from body by a blank line
  2. Subject is limited to 50 characters (not including Jira issue reference)
  3. Subject does not end with a period
  4. Subject uses the imperative mood ("add", not "adding")
  5. Body wraps at 72 characters
  6. Body explains "what" and "why", not "how"

Documentation

• In case of new functionality, my PR adds documentation that describes how to use it.
  • All the public functions and the classes in the PR contain docstrings that explain what it does
  • If you implement backwards incompatible changes, please leave a note in the Updating.md so we can assign it to a appropriate release

[ashb]

We should include the try_number in here too - (i.e if try 1 is still running somehow but we want to launch try 2 we don't want to monitor try 1 again.)

[dimberman]

Is that something that could happen? Would the execution date be the same?

[ashb]

Yeah, if the task fails and it retires :) (It's unlikely to still be running, but some error/odd behavoiur could make it happen, but the "key" that uniqueue identifies a Task Instance is (dag_id,task_id, execution_date,try_number)

[danccooper]

Hi guys,

I think we should consider the desired behaviour with this. If we're including the try_number then it opens a situation where multiple tries could be running alongside each other. I don't think that is ever desired & would certainly cause problems with my jobs (back to the original cause of this one 😄).

If we want to avoid monitoring the previous try, then should we check for it & kill it if exists before starting the next?

[AXington]

@danccooper @ashb Question about this, I think retry_number would actually be good here... Especially if you are debugging and [kubernetes][delete_worker_pods] is "False". Seems like it would delete the original pod under @danccooper's suggestion and you'd lose that for debugging purposes.

[dakov]

On our k8s cluster, we keep pods in the namespace for some time no matter the state they finished in (success, OOMKilled, evicted,....) -- I do not maintain the cluster and the decision is not up to me. Not having try_number as an identifying label of a pod makes airflow fail to launch another attempt.

[danccooper]

Hey @dakov I was going to suggest you just set the reattach_on_restart flag to False, however from checking the code I'm not sure this will work as expected which is probably the behaviour you're reporting? After discussing wth @kaxil he has rasied #10021 to cover this, suggest we continue discussion on there.

[ashb]

/cc @danccooper. We'll give you Co-Authored-By on this PR when we merge it.

[danccooper]

Thanks for moving this along @ashb @dimberman 👍

[codecov-io]

Codecov Report

Merging #6377 into master will decrease coverage by 0.13%.
The diff coverage is 78.84%.

@@            Coverage Diff             @@
##           master    #6377      +/-   ##
==========================================
- Coverage   80.59%   80.45%   -0.14%     
==========================================
  Files         626      626              
  Lines       36243    36271      +28     
==========================================
- Hits        29211    29183      -28     
- Misses       7032     7088      +56

Impacted Files	Coverage Δ
airflow/kubernetes/pod_launcher.py	91.97% <100%> (ø)	⬆️
airflow/executors/kubernetes_executor.py	58.19% <100%> (-0.81%)	⬇️
airflow/kubernetes/pod_generator.py	95.03% <100%> (+0.33%)	⬆️
...rflow/contrib/operators/kubernetes_pod_operator.py	88.29% <72.5%> (-10.21%)	⬇️
airflow/executors/sequential_executor.py	47.61% <0%> (-52.39%)	⬇️
airflow/utils/log/colored_log.py	81.81% <0%> (-11.37%)	⬇️
airflow/utils/sqlalchemy.py	86.44% <0%> (-6.78%)	⬇️
airflow/executors/__init__.py	63.26% <0%> (-4.09%)	⬇️
airflow/utils/dag_processing.py	56.23% <0%> (-2.67%)	⬇️
airflow/jobs/scheduler_job.py	73.72% <0%> (-1.21%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 74d2a0d...53c8469. Read the comment docs.

[kaxil]

@danccooper - FYI @dimberman is on holiday for a week, so I guess this will have to wait till he is back and has time to look at it.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

issues addressed