Warm up persistence infrastructure at startup

We got a report in which the first call to a PersistenceEntity may take a few seconds on newly deployed nodes.

This happens, particularly in CPU starved environments.

The ideal situation would be to trigger the setup of the persistence plugin at bootstrap. By the time that a node joins a cluster and receive the first real request, the plugin will be already configured.

Eventually, we could add a dummy persistent actor (no entity, not sharded), that we could hit as soon as possible. That actor won’t have any state or event, it only needs to implement PersistentActor and shutdown just after ‘recovery’.

[octonato]

reopening because #2370 was reverted in #2416

[howesro]

Hi, we are also facing similar issues here and it is particularly obvious in our staging environments.

Every time we release a Lagom Persistent Entity service the CI/CD pipeline fails on our automated test suite. Major red flag for me is when the team start discussing having our pipeline automatically run the tests twice to bypass. I think the suggestion to have the framework do this automatically makes a lot of sense - ultimately if the service wants to make use of Cassandra backed Persistent Entities then I see no downside to warming this up vs forcing every consumer to do the same.