network latency between nodes
sreeblr
- Select Field - Employee
intermittently one of the nodes in 3 node cluster on azure .goes down and then we restarted . Vertica.log shows spread messages. At the time when node goes down I am able to connect from one node to another without issue .Can we increase spread timeout ? Will any setting on vertica/Linux/azure help reduce the issue. We have red hat 7.2 ,Vertica 8.1 3 node cluster on azure.
with 018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] NETWORK change with 1 VS sets
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] VS set #0 (mine) has 1 members (offset=36)
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] VS set #0, member 0: #node_a#N143162100004
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] DB Group changed
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [VMPI] DistCall: Set current group members called with 1 members
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [VMPI] Removing 45035996273705106 from list of initialized nodes for session v_mydb_node0001-324324:0x26
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] NETWORK change with 1 VS sets
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] VS set #0 (mine) has 1 members (offset=36)
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] VS set #0, member 0: #node_a#N143162100004
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] DB Group changed
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [VMPI] DistCall: Set current group members called with 1 members
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [VMPI] Removing 45035996273705106 from list of initialized nodes for session v_mydb_node0001-324324:0x26
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [VMPI] Removing 45035996273705110 from list of initialized nodes for session v_mydb_node0001-324324:0x26
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Comms] nodeSetNotifier: node v_mydb_node0002 left the cluster
2018-03-23 14:41:06.233 Spread Client:7f44cad06700 [Recover] Node left cluster, reassessing k-safety...