[CASSANDRA-15645] Can't send schema pull request: node /A.B.C.D is down

• Type: Bug
• Status: Triage Needed
• Priority: Normal
• Resolution: Unresolved
• Component/s: Cluster/Schema
• Platform: All
• Assignee: Unassigned
• Reporter: Pierre Belanger apache.org
• Created: 16/Mar/20 17:08
• Updated: 20/Apr/20 20:34

Description

On a new cluster with Cassandra 3.11.5, each time a node joins the cluster the schema pull request happens before at least 1 node is confirmed up.  On the first node it's fine but node #2 and following are all complaining with below WARN.

 

INFO [MigrationStage:1] 2020-03-16 16:49:32,355 ColumnFamilyStore.java:426 - Initializing system_auth.roles
WARN [MigrationStage:1] 2020-03-16 16:49:32,368 MigrationTask.java:67 - Can't send schema pull request: node /A.B.C.D is down.
WARN [MigrationStage:1] 2020-03-16 16:49:32,369 MigrationTask.java:67 - Can't send schema pull request: node /A.B.C.D is down.
INFO [main] 2020-03-16 16:49:32,371 Gossiper.java:1780 - Waiting for gossip to settle...
INFO [GossipStage:1] 2020-03-16 16:49:32,493 Gossiper.java:1089 - InetAddress /A.B.C.D is now UP
INFO [HANDSHAKE-/10.205.45.19] 2020-03-16 16:49:32,545 OutboundTcpConnection.java:561 - Handshaking version with /A.B.C.D

 

It's not urgent to fix but the WARN create noise for no reason.  Before trying to pull the schema, shouldn't the process wait for gossip to have at least 1 node "up"?