Where is the best place to clean/refresh resources after queued jobs are handled?

[halaei]

Running queue:work command, I want to be sure that all the resources are clean and healthy before handling the next job. For example, I would like to rollback or reconnect the DB connection if the previous job was not successful in closing the transaction.

I was thinking to listen to Events\JobProcessing event. However, this event is also fired when pushing a job to a 'sync' queue, and I don't want to reset the DB connection every time I push a job to a 'sync' queue.

[sisve]

It sounds like you want to run the queue without the daemon mode. Running queue:work --once will only execute one job and then quit; this will close down all resources and database connections (as php scripts do at the end of execution). Running queue:listen will call queue:work --once repeatedly.

[halaei]

I am already aware of that option. However, queue:listen is not efficient enough for my use-case, because of the fork overhead, plus the unnecessary overhead of resetting all the remote connections, when there is no need to.