I'd like to raise some attention for the following 2 tickets, they're patch-ready and deployed on all our production clusters: * CASSANDRA-15098: "Endpoints no longer owning tokens are not removed for vnode" For vNode cluster, the replaced node may not be removed from gossiper (and system.peers, every time a node restart, they will be re-populated into gossiper from system.peers). The patch is pretty small and straightforward. * CASSANDRA-15097: "Avoid updating unchanged gossip state" I think this is a bug because it's not only causing large pending gossip tasks, it's also causing long token-metadata update lock for large vNode cluster.
Here is another improvement we made for vNode to avoid gossip block during removeNode: CASSANDRA-15141. But I think it can wait until 4.x. It mostly causes problem for 1000+ vNode clusters. Thanks, Jay On Tue, Jul 2, 2019 at 10:28 PM Mick Semb Wever <m...@apache.org> wrote: > > > > Is there any chance to get CASSANDRA-15005 (ready, with PR) into a > > 3.11.5 release? > > > I doubt it Soroka. It's not a bug and there's no patch for it, so I'd see > no reason why Michael would wait for this when he generously finds time to > cut a release. > > Maybe the author and reviewer decides to push it to both 3.11.x and 4.0, > but that's irrelevant to this thread. > > regards, > Mick > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@cassandra.apache.org > For additional commands, e-mail: dev-h...@cassandra.apache.org > >
