Promscale should not terminate under non-critical conditions

[mjf]

Promscale terminates with error under conditions when one would expect just some log entries. For example when it could not connect to a database (this may be because of many reasons and have simply nothing to do with Promscale as such, i.e. somebody may have deployed new Promscale instance and have not yet configured PostgreSQL role for Promscale for it, or even hasn't created the database yet, or there is pg_hba.conf entry missing etc.). I agree that software in general should terminate if there is a critical configuration error or serious runtime error but I do not agree it should terminate if the other side is just rejecting connections for any reason etc. So what I suggest is to make database connection errors non-critical so that normal connection retries apply. Optionally some option may be provided to influence the behaviour for backward compatibility.

[harkishen]

I agree we shouldn't terminate on all conditions. It would be better to categorize some as recoverable and some as non-recoverable and then terminal only on non-recoverable and the rest can be a retry.

[VineethReddy02]

closing this issue, as we haven’t received multiple requests from users for this feature and it's quite stale! :)

If you still have a need for this, Please open a new issue.

Thanks!