Hi all,
Could someone explain what are the possible downsides of using --skip-wait-for-gossip-to-settle 0
on a cluster with multiple nodes, if it should even be done?
I’m currently using it on a single node cluster (testing purposes) as it takes forever to start, is it normal? could it be caused by something else?
thanks
Once a node starts, all other endpoints are down. If you skip gossip, >=LOCAL_QUORUM
queries may fail if a coordinator query hits the RPC of that node.
Another issue: Make gossip synchronization on bootstrap more robust · Issue #2866 · scylladb/scylladb · GitHub
That said, you may skip it for testing, as indeed waiting ~30s at every startup can be annoying. Keep it as is on production tho.