postgres-10 pre-start script fails after upgrading Ops Manager to v2.10.40

Products

• Operations Manager

Issue/Introduction

If you have upgraded to Ops Manager 2.10.40 and your postgres-10 pre-start scripts fails when deploying Bosh Director then you are likely suffering from a known issue where remnants of the previous of postgres-9.4 installation were not properly cleaned up. Ops Manager 2.10.40 removed the migration logic for postgres-9.4 to postgres-10 which would have removed the postgres-9.4 directory. 

You will see the following error message in the postgres pre-start logs

Please use a previous bosh release version (271.x or lower) to migrate data from postgres-9.4 to postgres-10.

To confirm if you are suffering from this issue you will need to SSH into your Bosh Director VM.

1. Obtain Bosh Director BBR SSH Key and SSH into your Bosh Director
2. Open contents of file /var/vcap/sys/log/postgres-10/pre-start.stdout.log

This issue will be resolved in a future release of Ops Manager, please use the workaround below in the meantime.

Environment

Product Version: 2.10

Resolution

To workaround the issue you will need to remove the postgres-9.4 directory from the Bosh Director. WARNING: This directory should only be removed if your Bosh Director contains both the postgres-9.4 and postgres-10 directories. If only the postgres-9.4 exists then you will need to reach out to support.

1.  Obtain Bosh Director BBR SSH Key and SSH into your Bosh Director
2. Change directories to /var/vcap/store
3. Confirm if both directories (postgres-9.4 and postgres-10) exist after running ls -l
4. Remove postgres-9.4 directory sudo rm -rf /var/vcap/store/postgres-9.4
5. Rerun Apply Changes