kue-scheduler does not re-schedule unique job if worker dies uncleanly

[gmcnaught]

I believe this is due to a change I introduced:

If a unique job is 'active' and the worker that is processing the job becomes unavailable and never calls done(), and doesn't respond to the ttl timeout, the job will be stuck in 'active', and therefore never gets re-scheduled due to:

https://github.com/lykmapipo/kue-scheduler/blob/master/index.js#L56

    var isCompletedOrFailedJob =
      (job.state() === 'complete' ||
        job.state() === 'failed');

    if (isCompletedOrFailedJob) {

I will submit a pull request to also check if the last_updated value is older than the job ttl and schedule the next job - this way they do not end up stale.

The only reservation I have about this method is that in a situation where the worker is just unable to talk to redis, it may complete the work twice.

[lykmapipo]

appreciated

[TangMonk]

same problem

[bilby91]

I think we are having a similar issue here. After we received a timeout error from redis, the unique scheduled job never triggered again.

[Jageen]

I am having same problem and planning to do this

if Job is timeout
Create new job based on the job data.
Delete old job.

Is there any logical error on above flow ?