* A new wsrep provider option, gcache.recover, enables gcache to be recovered at restart. This allows the restarted node to serve IST to any other nodes that are also starting, which may speed up whole-cluster restarts. * An additional protection against restarting a cluster using a node that may not have all the transactions has been implemented. Unless a node was the last one to be shut down from the previous cluster, Galera will refuse to bootstrap from it until explicit action is taken by the operator. * Galera would refuse to start on PPC64EL (lp1633226, GAL-418) * Galera would crash or enter an infinite loop in the asio library (GAL-416) * It was not possible to set pc.wait_prim and pc.wait_prim_timeout options using wsrep_provider_options (GAL-360, codership/galera#360) * For security reasons, SSLv2, SSLv3, TLS 1.0 will no longer be negotiated during SSL handshake (codership/galera#428) * After a certain pattern of server restarts, an isolated node could stop attempting to reconnect to the cluster (GCF-890) * Galera would fail to compile with clang 3.8 (refs codership/galera#412) * The wsrep_desync_count variable could show a wrong value (GAL-401)
Just a blind copy seems to do the trick.
It has worked in production since november.
Bumped by Brian, thanks.