Read only when cluster has failed

[Michael Andrews]

Per google groups discussion https://groups.google.com/forum/#!topic/codership-team/FsdXXTUkLqQ I am submitting a feature request to support read only access to data on cluster nodes which have become isolated from the cluster or appear in a failed state due to issues with replication. This feature is critical for service providers who utilize mysql to provide data to applications that need 100% uptime for reads, regardless of the consistency of data on each individual node.

[Przemek (pmalkowski)]

This was already implemented in PXC by adding :
https://www.percona.com/doc/percona-xtradb-cluster/5.6/wsrep-system-index.html#wsrep_dirty_reads

[Michael Andrews (somecallmemike)]

Thanks for the comment. I am reviewing this variable and it appears this is a session scope? We need to globally configure all connections to support reading in the event the node is inaccessible. Is it possible to make this a global variable?

[Przemek (pmalkowski)]

Michael, it has also a global scope since PXC 5.6.26: https://www.percona.com/doc/percona-xtradb-cluster/5.6/release-notes/Percona-XtraDB-Cluster-5.6.26-25.12.html#new-features