TimeScaleDB - Cluster with failover and recoveryt.. Is it even possible
Hi All, I am running around in circles with the setup of TimeScaleDB. I have at the moment 2 nodes. 1 master, 1 slave.. For this example If reboot the master, it comes up as unresponsive. The slave becomes the master. However when the old master comes up. It stays as master in yellow. There is no option to promote back to master or even be a slave.. There is only the option to rebuild node to be a slave.
What I would be expecting is self recovery so if the new master fails, the slave will become master. Any recovered master returns to a slave or takes back it's master role after a database sync.
It seems timescaledb isn't setup for fail over and with a virtual IP to the current master.
Any suggestions please to assist in my setup?
-
Official comment
Hi Mark,
ClusterControl does not support that kind of automatic recovery for all master-slave replication setup (MySQL as well). This is due to the fact that every master failure is considered a catastrophic failure and has to be investigated by the user before let it join back to the cluster. There are possibilities where the old master would appear not fit to join the cluster even as slave after it's available like disk full, WAL logs have been rotated, etc.
Once the old master backs up, you have to manually sync it as a slave by going to CC -> Nodes -> pick the old master -> Node Actions -> Rebuild Replication Slave. Once it catches up, you can then promote this slave to be the master.
Please sign in to leave a comment.
Comments
1 comment