OpenNMS Crashing when we change PostgreSQL server IP and do helm Upgrade.

[gp185132-ncr]

Hi DJ,

In Production (Chart version 0.2.0 as we have upgrade Pending for Prod) we have noticed below behaviour:

Lets say we want to Switch Already existing OpenNMS deployed instance to completely new Postgres server (In our Case New Cloud SQL Postgres instance having new IP than already put IP under values.yaml field :
postgresql:
hostname: X.X.X.X

And if we do rollout restart of OpenNMS Statefulset , we observed that it crashes with errors , even we tried creating customer as username and customer_opennms as DB as OpenNMS do but still it fails , so we needed to completely uninstall the realease and deploy again . Is this the expected behaviour ? Can you check on this as knowing how to Migrate/shift Postgres Instance is important to us. I will share the necessary files over mail to you.

[deejgregor]

I suspect what you are running into is that the database for this instance on the new server is (almost) empty. And on version 0.2.0 of the chart, the installer isn't automatically executed again, so then the installer that is normally triggered by the chart won't be able to setup the database. Unfortunately, it's unsafe to remove the "configured" file to trigger the installer to run until later versions.

You'll need to transfer the database for that instance over to the new database server or run the installer outside of the Helm chart to initialize the database.

With the helm chart >= 0.4.0, you'll be able to remove the "configured" file and it will run the installer again.

Does this explanation help?