TypeError: can't pickle _thread.lock objects when cleaning about airflow-provider-great-expectations

[edgarrmondragon]

@Casyfill Oh I think I just found a solution 😅. Threads seem to be used to send usage statistics so I disabled them in great_expectations.yml:

anonymous_usage_statistics:
  enabled: false

[spbail]

Thanks for helping triage this! Obviously this isn't an ideal solution form our perspective since we'd like to keep usage stats turned on if possible, but I'm glad there's at least a workaround. I'll tag this for review and maybe we can come up with a permanent fix.

[edgarrmondragon]

We're on Airflow 1.10.13 and running into the same issue when trying to clear a GreatExpectationsOperator task.

[Casyfill]

Thanks @edgarrmondragon , that makes a lot of sense.
Perhaps still makes sense to turn this off automatically (or flash a warning) by default as part of the operator

[edgarrmondragon]

Perhaps still makes sense to turn this off automatically (or flash a warning) by default as part of the operator

That would be reasonable as the true solution.

[Casyfill]

2 cents - it is obviously dependent on your deployment, but apparently, if you remove data_context_id (which I assumed to be unused given enabled:false) from the expectations yaml, it will attempt to override this YAML upon initialization. To avoid that (which does not make sense in airflow, and is forbidden by our settings), just get this line back, while keeping enabled:false

[scoyne2]

On airflow-provider-great-expectations==0.0.5 and great-expectations==0.13.9 I too get error can't pickle _thread.lock objects when attempting to clear the GE Task using the below.
(I am using Astronomer with Airflow 1.10.7

anonymous_usage_statistics:
  enabled: false

If I downgrade to airflow-provider-great-expectations==0.0.4 and great-expectations==0.13.7 I do not see this issue (except when attempting to use a Postgres MetricStore