Removal of Async replication
WARNING: it is an ‘8.0/candidate’ article. Do NOT use it in production!
Contact Canonical Data Platform team if you are interested in the topic.
Pre-requisits
Make sure both Rome and Lisbon Clusters are deployed using the Async Deployment manual!
Detach Cluster from ClusterSet
Note: It is important to switchover the
PrimaryCluster before detaching it from ClusterSet!
Assuming the Lisbon is a current Primary and we want to detach Rome (for removal or reuse):
juju remove-relation async-primary db2:async-replica
The command above will move cluster Rome into the detached state blocked keeping all the data in place.
All units in Rome will be in a standalone (non-clusterized) read-only state.
From this points, there are three options, as described in the following sections.
Rejoin detached cluster into previous ClusterSet
At this stage, the detached/blocked cluster Rome can re-join the previous ClusterSet by restoring async integration/relation:
juju switch rome
juju integrate async-primary db1:async-replica
Removing detached cluster
Remove no-longer necessary Cluster Rome (and destroy storage if Rome data is no longer necessary):
juju remove-application db1 # --destroy-storage
New ClusterSet from detached Cluster
Convert Rome to the new Cluster/ClusterSet keeping the current data in use:
juju run -m rome db1/leader recreate-cluster