No, it’s not possible to have Oban use a read replica for standard operation (running jobs), because fetching and executing jobs is a write operation. Fetching records that an attempt happened, where it happened, and at what time. Then, when execution finishes, it records the status. All of those are write operations.

You’ll need to handle that situation in your workers or by scheduling jobs slightly in the future to compensate for the lag.

Thanks, yes already handling it on the job side. Tried the scheduled_in option but due to timezone issues it didn’t quite workout. We are good for now and thanks for such an amazing piece of tech.

Of note for future readers, the schedule_in and scheduled_at timestamps are always UTC. Oban only uses time zones for cron schedules