Restart RabbitMQ consumers when vital queue got no consumers

[quantranhong1999]

Why

Today, there were no consumers on deleted-message-vault-work-queue of MU, so we needed to restart TMail ourselves.

We should have a mechanism to automatically spot the no consumers issues and restart them.

Originally it seems @chibenwa thought about writing a healthcheck for RabbitMQ consumers on the queues + auto restart TMail.

However, I checked the code and got another simple idea.

How

• Add the deleted-message-vault-work-queue (could be other queues as well) to ScheduledReconnectionHandler monitoring list

DoD

• Test the ScheduledReconnectionHandler more to ensure it can correctly restart the consumers if a queue in the list has no consumer.

[chibenwa]

Deleted Message Vault may be disabled.

However the list of queues is static. We need to make sure queue consumption still runs if ScheduledReconnectionHandler has queue configured that do not exist

(write explicit tests for it, passing abc-i-do-not-exist as a queue in tests?)

[quantranhong1999]

(write explicit tests for it, passing abc-i-do-not-exist as a queue in tests?)

+1

However the list of queues is static.

We can make the list configurable.

[quantranhong1999]

Likely we need a ScheduledReconnectionHandler implementation for deleted message vault work queue on James side.

[chibenwa]

No my feeling is that by handling the missing queue leniently we could keep this sttatic, just monitor the queue that exist and keep the configuration burdown low.

[chibenwa]

I'm against conf for this: it will get forgotten. The alternative would be to link it with the module chooser but it looks complex...

[quantranhong1999]

Likely we need a ScheduledReconnectionHandler implementation for deleted message vault work queue on James side.

apache/james-project#2546

(write explicit tests for it, passing abc-i-do-not-exist as a queue in tests?)

#1366