Job with multiple executions #1964
Conversation
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #1964 +/- ##
==========================================
+ Coverage 93.61% 93.84% +0.22%
==========================================
Files 28 29 +1
Lines 3757 3897 +140
==========================================
+ Hits 3517 3657 +140
Misses 240 240
☔ View full report in Codecov by Sentry. |
| @@ -55,6 +56,13 @@ class JobStatus(str, Enum): | |||
| CANCELED = 'canceled' | |||
|
|
|||
|
|
|||
| def parse_job_id(job_or_execution_id: str) -> str: | |||
| """Parse a string and returns job ID. This function supports both job ID and execution composite key.""" | |||
| if ':' in job_or_execution_id: | |||
There was a problem hiding this comment.
This is a problem: when creating a standard job, you can pass a job_id to it, so that instead of using the default uuid, you can use your own custom job id convention.
We use that, and we already use : and | as special separator in the job id. This code change breaks the tacit contract and now a job id can't contain some characters. This could break stuff in production for some users (it will for us...)
Can't the execution id be stored somewhere else ?
Since rq/rq#1964 the job status is set to Failed before the handler decides whether to capture or not the exception while handle_job_failure has not yet been called so the job is not yet re-scheduled leading to all exceptions getting captured in RQ version >= 2.0. Related to #1076 Fixes #3707
This PR introduces an
Executionobject to represent a job's execution. Completion of this PR should allow us to properly track multiple job executions.