pg_hba.conf does not match postgres image

[belak]

This means that anything not connecting via the local socket will fail, so most applications linking the postgres container will not work.

At least for me, pg_hba.conf was missing this:

host all all 0.0.0.0/0 md5

Normally this would be set up in https://github.com/docker-library/postgres/blob/master/9.5/docker-entrypoint.sh, but because we use initdb during the upgrade, that never happens.

[tianon]

Crap, looks like we need to shell out to "docker-entrypoint.sh" for initdb somehow, which might actually require something like "docker-initdb" to be created from just those initdb lines. A workaround would be to run the image you're trying to move to on your new data directory first so that the real initdb is run, then stop it, and run the upgrade image.

[CvX]

@tianon I've changed db setup so it looks like this: master...CvX:new-db-setup-fix#diff-cd39faa12723607353bf74d281b3ed73
That works for me. Are there any downsides to running postgres' docker-entrypoint.sh just like that?

[tianon]

Future changes to that script could technically break it, but it's probably generally safe. 👍

[WalkerWalker]

just realized that the pg_hba.conf file is different in the data folder of two postgres versions
In my old postgres, The last linehost all all all md5 exists, but in my new postgres, this line does not exist.
Then I found this issue. So is it still open?