Postgres: use pg_try_advisory_lock instead of pg_advisory_lock

[AkuSilvenius]

Issue

#960

Changes

• For postgres Lock, use pg_try_advisory_lock instead of pg_advisory_lock to fix the expected behavior of waiting for acquiring lock indefinitely
• Add config parameter x-lock-retry-max-interval to configure the max interval to be used for exponential backoff when retrying Lock

[coveralls]

coverage: 56.454% (+0.1%) from 56.335%
when pulling 49bee23 on AkuSilvenius:960_add_try_advisory_lock
into 0c456c4 on golang-migrate:master.

[dhui]

left a comment in case we decide to go this route

[dhui]

Use the backoff library with an exponential backoff + jitter as the default. Note, this may cause other nodes/hosts to take longer to deploy due to the longer wait period so the backoff should be configurable.

[AkuSilvenius]

@dhui thanks for the feedback, I've added a configurable exponential retry.

this may cause other nodes/hosts to take longer to deploy due to the longer wait period

I've added defaults and kept them quite small to address the above

[AkuSilvenius]

unfortunately the tests keep failing because device is running out of space

[eelco]

I assume the tests were temporarily broken, so this might work after a rebase? I’m interested in this fix landing 😊

[AkuSilvenius]

I assume the tests were temporarily broken, so this might work after a rebase? I’m interested in this fix landing 😊

@eelco thanks for noting, looks like #1072 fixed the tests and pipeline here in the PR is working and passing now - unfortunately this PR got stuck in sleep mode for quite a while due to the unstable pipeline, I hope we can merge this soon :)