Successful RQ job execution does not get removed from `StartedJobRegistry`

[choo0166]

I have defined a job using the job decorator syntax like:

@job("calcium", timeout=3600, on_success=report_success, on_failure=report_failure)
def my_task(payload):
    start = time.perf_counter()
    response = requests.post(
        url,
        headers={
            "Content-Type": "application/json",
        },
        json=payload,
    )
    response.raise_for_status()
    end = time.perf_counter()

    response_body = response.json()

    return response_body

I have also defined the success callback handler:

def report_success(job, connection, result, *args, **kwargs):
    try:
        analysis = Analysis.objects.get(id=job.id)
        analysis.status = Analysis.Status.COMPLETED
        if "prediction" in result:
            PredictionResult.objects.create(
                analysis=analysis, prediction=result["prediction"]
            )
        if "segmentation" in result:
            # handle decoding and storing of base64 encoded segmentation masks
            for tp, (name, data) in result["segmentation"].items():
                # decode base64 string into binary and save to filefield
                f = ContentFile(content=base64.b64decode(data), name=name)
                SegmentationResult.objects.create(analysis=analysis, segmentation_mask=f, mask_type=tp)
        if "artifacts" in result:
            # handle decoding and storing of intermediate model artifacts
            for tp, (name, data) in result["artifacts"].items():
                f = ContentFile(content=base64.b64decode(data), name=name)
                AnalysisArtifact.objects.create(analysis=analysis, artifact=f, artifact_type=tp)
        analysis.save()
        logger.info(f"Analysis {analysis.id} processed successfully")
    except Analysis.DoesNotExist:
        logger.error(f"Analysis with job id {job.id} not found")

Basically, it handles saving of base64 encoded data into the respective Django model filefields.

When the job execution completes, RQ fails to remove the job from the StartedJobRegistry:

ai_worker-1     | Analysis 8df26860-bdcc-49da-a537-43088e7bddca processed successfully
ai_worker-1     | Handling successful execution of job 8df26860-bdcc-49da-a537-43088e7bddca
ai_worker-1     | Saving job 8df26860-bdcc-49da-a537-43088e7bddca's successful execution result
ai_worker-1     | Removing job 8df26860-bdcc-49da-a537-43088e7bddca from StartedJobRegistry
ai_worker-1     | Saving job 8df26860-bdcc-49da-a537-43088e7bddca's successful execution result
ai_worker-1     | Sent heartbeat to prevent worker timeout. Next one should arrive in 90 seconds.
ai_worker-1     | Removing job 8df26860-bdcc-49da-a537-43088e7bddca from StartedJobRegistry
ai_worker-1     | Saving job 8df26860-bdcc-49da-a537-43088e7bddca's successful execution result

The last two log lines basically repeat themselves till the job timeout is reached.

Any ideas to debug this issue?

[choo0166]

I've found that it was the size of the result json that was too large to be saved in redis, setting result_ttl=0 in the job decorator is a workaround for this issue. Is there any way to specify an option to not save the results in redis, but save the job execution detail instead?