WIP: refactor: reduce code duplication in request processing #892
Conversation
This provides the agreed upon distinction between
- PENDING
a normal occurrence of a request not yet running
- PENDING_WORKERS_OFFLINE
an abnormal occurrence where the request has not yet run and cannot run due
to lack of workers; note that this includes auto-start events where we
definitively know the request's state
- WORKERS_OFFLINE
an abnormal occurrence where we do not know the request's state
PENDING_WORKERS_OFFLINE maintains the fail-over functionality previously dependent
solely on WORKERS_OFFLINE.
For consistency with other log messages from the processing code the content of the message indicating issues with the workers has been modified from that documented in AAP-23378. It is worded as follows:
The workers available to process request type {request_type} for
{process_parent_type} {process_parent_id} are failing liveness checks.
There may be an issue with the node; please contact the administrator.
Orchestrator log messages were updated due to previous change that specifically set
the request type if none was specified.
|
To me that message is contradictory - it implies there are workers available, when in fact they are not responding to liveliness checks, which in essence means they are unavailable. |
* bump DAB version to 2024.5.6 * add PATCH action for detail view OPTIONS * add tests for PATCH/POST actions on OPTIONS request Co-authored-by: Alan Rominger <[email protected]>
@ttuffin Updated per our slack convo. |
* add a custom "sync_project" permission for project sync action
|
All outstanding issues and critiques have been addressed. |
|
HI @jshimkus-rh I'm finally covered this with #894 |
jshimkus-rh
changed the title
True. However, I saw that as a stopgap to fixing the regression from #880. |
|
I'm addressing some other issues there as well. Also, I'm not sure, but I had there to update some tests, which makes sense since we are changing the behavior. I'm receiving some "pressure" from mates because right now the devel builds are mostly in an useless state, I'm trying to finish it as soon as possible to unblock it and continue working on other issues. I would like to take advantage of the situation to extend the coverage of the integration tests, but due to the rush for the fix I think we can do it in later PR's. I suggest to continue working there, because I'm already tested it in different deployments and try to have it ready for merge soon. |
|
Converted to draft. To be revisited post next release. |
jshimkus-rh
changed the title
This provides the agreed upon usage of PENDING w/ status text containing the specific message:
as distinguishing a request that cannot be run due to complete lack of workers.
The message text above is slightly modified from that documented in AAP-23378 for consistency with other log messages from the processing code.
Orchestrator log messages were updated due to previous change that specifically set the request type if none was specified.