Jelte Fennema-Nio

I figured out why I was getting `localhost` in there. It was because I had `export PGHOST=localhost` in my `~/.bashrc` file. Without this `pg_autoctl` worked fine when doing `--auth-host=reject --auth-local=trust`....

> We seem to have `create monitor --nodename` parameter. Do you have something different in mind @JelteF? For `create postgres` we have both a `--nodename` and a `--pghost` parameter. `--nodename`...

I understand what you mean, which is why I put BONUS there originally. The main problem it causes is that when doing `create monitor` with `PGHOST=localhost` without `--auth-host=trust` then the...

> - ignore PGHOST and skip implementing `--pghost` on the monitor, Fine by me > - remove and replace the `pg_hba.conf` file wholesale after `pg_ctl initdb` to have our own...

I agree that that's the case when the database already exists. But if we create it we should not open it up by default when `pg_autoctl` does not need it.

I understand what you mean, but I disagree. The default for postgres is to warn about it's insecure default, to try to get the user to change it. We actively...

My main point is, we're already adding a lot of pg_hba.conf rules anyway also for `create postgres`. So making them a bit secure by default seems like a nice thing...

> In the future, we might consider adding --auth-local and --auth-host parameters to create postgres subcommand like here, but for now, we thought lets leave is as is. If we...

Hmm that looks like a bad situation indeed and this might be a bug in pg_auto_failover. This part of the log tells the problem: ``` Apr 28 22:06:19 node_15 pg_autoctl[3514691]:...

Also if this is a prod cluster and you just want to recover as soon as possible, the easiest would be to drop the node that is stuck from the...