卡桑德拉 - 种子八卦版本是-2147483648

Question

I've just noticed a weird scenario in my UAT environment.

I've got a 3 node cluster but I noticed this morning that nodes 2 and 3 think node 1 is dead. Node 1 however thinks everyone is alive.

In the logs for 2 and 3 it says the following:

WARN  [MessagingService-Outgoing-/10.0.8.172] 2015-12-06 02:20:02,987 OutboundTcpConnection.java:423 - Seed gossip version is -2147483648; will not connect with that version

Also it appears node 1 is no longer listening on 9042. It is still listening on 7000 though.

worth noting I'm on Windows Server 2008 R2 and running Cassandra 2.2

Thanks

Answer 1

Do node 2 and 3 know that node 1 exists and that it is down or do they not know that it exists at all?

Have you checked the settings in your cassandra.yaml file? Are you sure the windows firewall on node 1 is not blocking the ports? Look here: https://docs.datastax.com/en/cassandra/2.0/cassandra/security/secureFireWall_r.html

If they know that it exists but they think that it is down then I would look in the cassandra.yaml file and specifically look at the listen_address on node 1.

Answer 2

To solve this, just set:

Node 1: seeds = node1,node2

Node 2: seeds = node2.

---

1st start node 1 and then node 2!

Fixed my problem!