repmgr prevents postgresql upgrade?

[lordcirth]

I am trying to upgrade a prototype postgresql 12 master to 13. The host is Ubuntu 18.04 with both 12 and 13 postgresql + repmgr installed. When running pg_upgradecluster without -m upgrade, it just hangs. With -m upgrade, it fails with an error related to repmgr's database.

pg_upgradecluster 12 main
pg_upgradecluster 12 main -m upgrade

In pg_upgrade_server.log:

2020-12-08 16:41:21.665 EST [78369] postgres@repmgr STATEMENT:  CREATE FUNCTION "repmgr"."am_bdr_failover_handler"(integer) RETURNS boolean
            LANGUAGE "c" STRICT
            AS '$libdir/repmgr', 'am_bdr_failover_handler';
        
        -- For binary upgrade, handle extension membership the hard way
        ALTER EXTENSION "repmgr" ADD FUNCTION "repmgr"."am_bdr_failover_handler"(integer);

postgres@mc-3015-postgres-1804:/var/log/postgresql/pg_upgradecluster-12-13-main.te8P$ dpkg -l | grep postgres
ii  pgdg-keyring                              2018.2                              all          keyring for apt.postgresql.org
ii  postgresql-12                             12.5-1.pgdg18.04+1                  amd64        object-relational SQL database, version 12 server
ii  postgresql-12-repmgr                      5.2.0-2.pgdg18.04+1                 amd64        replication manager for PostgreSQL 12
ii  postgresql-13                             13.1-1.pgdg18.04+1                  amd64        object-relational SQL database, version 13 server
ii  postgresql-13-repmgr                      5.2.0-2.pgdg18.04+1                 amd64        replication manager for PostgreSQL 13
ii  postgresql-client-12                      12.5-1.pgdg18.04+1                  amd64        front-end programs for PostgreSQL 12
ii  postgresql-client-13                      13.1-1.pgdg18.04+1                  amd64        front-end programs for PostgreSQL 13
ii  postgresql-client-common                  223.pgdg18.04+1                     all          manager for multiple PostgreSQL client versions
ii  postgresql-common                         223.pgdg18.04+1                     all          PostgreSQL database-cluster manager

Full logs:
pg_upgradecluster-12-13-main.tar.gz

I am not very familiar with postgresql or repmgr, so my mistake might be obvious. Thanks.

[ibarwick]

See the relevant documentation section: https://repmgr.org/docs/current/upgrading-and-pg-upgrade.html

Basically you should upgrade repmgr on the old cluster before running pg_upgrade on PostgreSQL itself.

[lordcirth]

Thanks. I knew I was forgetting something; I did try to upgrade repmgr but couldn't. I updated the apt packages to 5.2, but ran into this error:

postgres@mc-3015-postgres-1804:~$ repmgr -f /etc/repmgr.conf primary register
INFO: connecting to primary database...
ERROR: an older version of the "repmgr" extension is installed
DETAIL: version 5.0 is installed but newer version 5.2 is available
HINT: update the installed extension version by executing "ALTER EXTENSION repmgr UPDATE"

postgres=# ALTER EXTENSION repmgr UPDATE;
ERROR:  extension "repmgr" does not exist
postgres=# 

But just now I realized that the extension is database-specific, not global, so \c repmgr was required to update the extension:

postgres=# \c repmgr
repmgr=# ALTER EXTENSION repmgr UPDATE;
ALTER EXTENSION

Perhaps this is obvious to others, but the HINT was not clear to me.

postgres@mc-3015-postgres-1804:~$ pg_upgradecluster 12 main -m upgrade appears to have succeeded now. Thank you for your help!