fix(deployment): Postgres healthcheck, add username to pg_isready

[mmomjian]

I believe this is the cause of the FATAL: role "root" does not exist error that has been popping up. In my testing pg_isready is available to all users, but that might not be the case in all installations.

Might fix #10219

[bo0tzz]

From https://pgpedia.info/p/pg_isready.html I understand that the db and user flags aren't even really relevant, so we shouldn't need to set them. I'm fine with whatever though :P

[mmomjian]

From https://pgpedia.info/p/pg_isready.html I understand that the db and user flags aren't even really relevant, so we shouldn't need to set them. I'm fine with whatever though :P

This explains why it still worked in testing. That wiki page states: “ However the server will log the validity of the connection attempt.”

FATAL: role "nosuchuser" does not exist

[bo0tzz]

Yeah, I think that error log on #10219 is just a red herring and we don't actually need this change.

[mmomjian]

Yeah, I think that error log on #10219 is just a red herring and we don't actually need this change.

The pg_pedia wiki page you linked explains this error exactly, doesn’t it? It says if a non existing user (root in this case) connects, it will still get a status 0 ready result, but PG will log it as a FATAL error. That doesn’t seem good.

[bo0tzz]

The healthcheck succeeds. Postgres logs a spooky error, but that isn't actually a hard problem, it just doesn't look very nice.

[mmomjian]

Oh yes, I agree with you there. I thought you were saying we shouldn’t make this change, but I guess you meant it’s not the true underlying source of that other issue?

[bo0tzz]

Both, but with a very weak should - I think this change is fine but probably unnecessary.