Connection lifetime

[metala]

Rationale.
In order to mitigate the issue of long running connections that increase database memory usage over time I have introduced lifetime to all connections of the pool. I need your feedback on the patch.

[iby]

@metala Hey there! Just curious if this attempts to solve #163 and removes hanging connections after completed queries?

[metala]

@iby Actually, it's about an OOM issue I hit with PostgreSQL as one of the services was making a lot of specific queries and crashing the database server. It might work for that case as well, but the connection lifetime is set using time duration and not number of requests.

The issue is about how Linux's fork() and memory CoW work. It is better explained here: https://italux.medium.com/postgresql-out-of-memory-3fc1105446d

[porsager]

Yeah, #163 didn't really have an issue did it? what are you having issues with @iby ? (maybe open a new issue)

[metala]

I am trying to run all the tests, but the test runner is broken on my setup. I have just ensured that my new tests are running fine.

My test setup:

$ docker run -it -e 'POSTGRES_USER=metala' -e 'POSTGRES_DATABASE=postgres_js_test' -e 'POSTGRES_HOST_AUTH_METHOD=trust' -v '/var/run/postgresql/:/var/run/postgresql/' postgres:13
$ yarn test

Edit. I have some new thoughts.

1. Maybe the slot should be released only when the connection has ended. Thus we will not exceed pool max connections.
2. Rename release to gracefullyEnd and it's flag from released to endingGracefully.

Let me know what you think about that.

[metala]

We are switching DB, so I don't think I would be able to maintain my change here, thus I will just close the MR.