I have to manage a project where a cluster of 3 nodes has been setup. I ran on every node “SHOW GLOBAL STATUS LIKE 'wsrep_%” to see about replication staus. Two nodes (node1 & node3) return the same fields but node2 shows more rows, and have a different
present on node 2, but absent on node 1 and 3
wsrep_local_send_queue_max
wsrep_local_send_queue_min
wsrep_local_recv_queue_max
wsrep_local_recv_queue_min
wsrep_cert_bucket_count
wsrep_gcache_pool_size
wsrep_cert_interval
wsrep_evs_delayed
wsrep_evs_evict_list
wsrep_evs_repl_latency
wsrep_evs_state
wsrep_gcomm_uuid
wsrep_local_index node 1 and 3 : 1 node 2 : 2
wsrep_provider_version node 1 and 3 : 3.4(r176) node 2 : 3.14(r53b88eb)
I then used the percona toolkit to compare config on the 3 nodes and again node2 shows some differences
14 config differences
Variable mysql01 mysql03 mysql02
========================= ================= ================= ================
auto_increment_offset 2 1 3
base_host 172.20.64.134 172.20.64.136 172.20.64.135
evs.install_timeout PT15S PT15S PT7.5S
evs.max_install_timeouts 1 1 3
hostname mysql01 mysql03 mysql02
repl.proto_max 5 5 7
My questions:
- is it a normal situation ? I guess no… Should I immediately disconnect node 2 ?
- should I evict the node2 from the cluster and setup a new node or is there a way to ‘replicate/synchronize’ the config on the 3 nodes ?