Using DbScheduler Without Continuous Polling

[francescods]

Hi,

I have a question regarding the possibility of using DbScheduler without continuous polling in a Spring Boot project.

My use case is as follows:
I only need to schedule one-time tasks sporadically, typically with a delay of around 30 seconds.
I would like the polling interval (e.g., 1 second) to be active only when there are tasks waiting for execution.
However, when there are no tasks pending, I would like to disable or reduce polling to avoid unnecessary database queries.
Is this behavior possible with the current configuration options? If so, could you provide some guidance on how to achieve it?

Thanks in advance for your time and help!

[eijckron]

I was also wondering about something similar but using a dynamic poll based on the known schedule.
Assuming the scheduler knows the next scheduled execution and can be triggered if new executions are added it knows the exact time when a new execution is planned and could therefore just skip polling until that time.
This can be messed up by concurrency but worst case scenario would be that an instance polls for work that is already picked up by another instance.
I assume it would still need an occasional poll (once every couple of minutes) to pick up work if another instance has stopped but it would be possible to have mostly seconds scheduling precision with a low polling frequency.

[kagkarlsson]

It is currently not possible, but I have entertained the idea of some sort of look-ahead mechanism for better precision. That will in itself be a form of polling though, but might get away with lower polling-frequency using such a strategy.

Another thing is that there is an optimization currently, that if a scheduler schedules an execution to now(), that scheduler will notify itself to go check for new executions directly. One might build upon that for this feature..

[francescods]

Hi,
ultimately, we opted to use it with continuous polling since we're now using db-scheduler for other needs as well, and we're very satisfied with this implementation. In any case, these seem like excellent proposals, thanks for the replies.