data store: fix incorrect task state

[wxtim]

…tidied

Closes #5645

Caused by a change in #5231

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Applied any dependency changes to both setup.cfg (and conda-environment.yml if present).
• Tests are not included because any test for this would be excessively specific.
• CHANGES.md entry included if this is a change that can affect users
• Docs not required
• If this is a bug fix, PR should be raised against the relevant ?.?.x branch.

[oliver-sanders]

@dwsutherland any thoughts on how the has_updated variable should play with this data store method?

[dwsutherland]

@dwsutherland any thoughts on how the has_updated variable should play with this data store method?

I think is because the update_data_structure method also includes non-data-store processing:

        # Database update
        self.workflow_db_mgr.put_task_pool(self.pool)
        self.update_data_store()

[dwsutherland]

Think I reproduced this problem.. by re-triggering a succeeded task:

(stuck in running state)

Fixed just doing this:

(flow) sutherlander@cortex:cylc-flow$ git diff

diff --git a/cylc/flow/scheduler.py b/cylc/flow/scheduler.py

index 3e1b40403..20c6c2d4d 100644

--- a/cylc/flow/scheduler.py

+++ b/cylc/flow/scheduler.py

@@ -1755,7 +1755,7 @@ class Scheduler:

                     self.timers[self.EVENT_RESTART_TIMEOUT].stop()

                     self.is_restart_timeout_wait = False

 

-            if has_updated:

+            if has_updated or self.data_store_mgr.updates_pending:

                 # Update the datastore.

                 await self.update_data_structure(self.is_reloaded)

 

@@ -1860,7 +1860,6 @@ class Scheduler:

                 sleep(0)

         # Database update

         self.workflow_db_mgr.put_task_pool(self.pool)

-        self.update_data_store()

 

     def check_workflow_timers(self):

         """Check timers, and abort or run event handlers as configured."""

(appeared to anyway)

[dwsutherland]

But I think you have it now.. will review.

[dwsutherland]

Probably just remove that self.update_data_store() call under has_updated ..
This actually may have effected the state-totals (..etc), i.e. #5595 (but I have a PR for that)

[dwsutherland]

This problem was quite pernicious, I think it made some tests flakier too (by creating a race condition with state updates .. i.e. tests/f/cylc-show/06-past-present-future.t)

[dwsutherland]

LGTM 👍

[oliver-sanders]

LGTM:

• Reproduced the original issue, confirmed this fixed it.
• Jammed print statements in to check that the data store update methods were being called as expected.