Fencing 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!
Fencing
In case of emergency failover, and there is a risk of the transaction sets differing between parts of the ClusterSet. To avoid a split brain scenario, where more than one cluster is set as primary, it’s important to fence all write traffic from the failed primary cluster. For doing so there’s an action fence-writes:
juju run -m rome db1/leader fence-writes cluster-set-name=<my-cluster-set>
where cluster-set-name is a mandatory option to avoid human mistakes (see below).
Note: The action
fence-writescan be run against any of the Primary Cluster units.
Unfencing
Case the old primary is reestablished and/or have all transactions reconciled, one can resume write traffic to it, by using the unfence-writes action, e.g.:
juju run -m rome db1/leader unfence-writes cluster-set-name=<my-cluster-set>
where cluster-set-name is a mandatory option to avoid human mistakes (see below).
ClusterSet name:
The cluster-set-name can be set on deployment and retrieved using:
juju run db1/0 get-cluster-status cluster-set=true
...
domainname: cluster-set-bcba09a4d4feb2327fd6f8b0f4ac7a2c
...
Extra
You can find more details about fencing in official MySQL ClusterSet documentation.