I want to build a Postgresql Cluster in a Cloud environment (yes I know it's a bad idea ... but that is not the question). For your information, I use a Dockerized Postgresql (see: https://github.com/paunin/postgres-docker-cluster) but that is not the problem and not the question.
So I have installed 2 Postgresql 9.5 with streaming replication. On each host I have repmgr 3.2.1 installed and correctly configured.
The following case have some issue:
- stop the master,
- wait for failover on slave -> ok,
- restart the old master -> Ko I've got 2 master in the cluster showed by
repmgr cluster showand I don't find any solution to force the old master to become slave.
The only ugly solution I found is to remove all data of the old master and restart it. Doing this, it became slave as expected (but a full resynchro is needed...).
Please note that It seems to be not possible to force the old master to became slave. repmgr standby register don't work as it says:
[2016-12-06 15:04:04] [INFO] connecting to standby database
[2016-12-06 15:04:04] [ERROR] this node should be a standby (user=replication_user password=replication_pass host=psg-host21 dbname=replication_db port=5432 connect_timeout=2)
[2016-12-06 15:04:04] [INFO] connecting to database
Role | Name | Upstream | Connection String
----------+------------|------------|------------------------------------------------------------------------------------------------------------------
* master | psg-host21 | | user=replication_user password=replication_pass host=psg-host21 dbname=replication_db port=5432 connect_timeout=2
* master | psg-host22 | psg-host21 | user=replication_user password=replication_pass host=psg-host22 dbname=replication_db port=5432 connect_timeout=2
Any help would be greatly appreciated.
