Though the option to retry failed instances of a publication has been around for sometime now, there are still some confusions around this option.
If you right click on any of the failed instances of a Publication, you will find three options
While other options are well documented, "retry" is not still very clear
Overwrites the "failed" instance (run now and reschedule will create new instances, but retry will use the failed instance itself)
In case of partial failure – retry option will process only the failed recipients.
In case of complete failure – the full job runs and is same as run now option- except for the fact that a new instance is NOT created when we retry.
In case if the server stops abruptly (example, you try to force restart SIA or the full box), the progress is not saved and so when the server comes up again, the instance that was running while the server was shutdown will be restarted from the beginning.
We can automate it using the “number of retries allowed” under the “recurrence” property of the publication.
In case of a failure, it will wait for the specified duration and then will attempt to run the publication again.