[jira] [Created] (CASSANDRA-12948) Upgrade 2.2.4 to 3.0.10 failed

["Romain (JIRA)" <ji...@apache.org>]

Romain created CASSANDRA-12948:
----------------------------------

             Summary: Upgrade 2.2.4 to 3.0.10 failed
                 Key: CASSANDRA-12948
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12948
             Project: Cassandra
          Issue Type: Bug
          Components: Core
         Environment: CentOS7, dsc-cassandra 2.2.4 / 3.0.10
            Reporter: Romain
            Priority: Blocker


Hi guys,

I have a problem when I try to migrate my cluster 2.2.4 to 3.0.10.

In production environnement, I don't want to stop my cluster, so I'll have mixing cluster during the migration.

Here is the cluster description in development and my tests:

Today, I have three nodes in 2.2.4, each one with 256 vnodes.

When I spawn a blank node 2.2.4, my keyspaces are well propagate to the new node.

When I spawn a blank node 3.0.10, my keyspaces are not propagate to the new node.
After the new node was spawned my application generate this error on the new node:

{code}
WARN  17:08:14 UnknownColumnFamilyException reading from socket; closing
cassandra3_1            | org.apache.cassandra.db.UnknownColumnFamilyException: Got legacy paged range command for nonexistent table docker.user.
{code}

It seems normal because keyspaces are not propagate.

Here is the output of the describecluster command:

{code}
Cluster Information:
	Name: Test Cluster
	Snitch: org.apache.cassandra.locator.DynamicEndpointSnitch
	Partitioner: org.apache.cassandra.dht.Murmur3Partitioner
	Schema versions:
		a34bf9e6-9139-3696-b85e-3fdd47aca08f: [172.17.0.2, 172.17.0.3, 172.17.0.5]

		59adb24e-f3cd-3e02-97f0-5b395827453f: [172.17.0.4]
{code}

Now, a new test. 3 nodes in 2.2.4, 0 in 3.0.10

I upgrade an existing 2.2.4 with this instruction https://docs.datastax.com/en/upgrade/doc/upgrade/cassandra/upgrdCassandraDetails.html.

So I have this describecluster:

{code}
Cluster Information:
	Name: Test Cluster
	Snitch: org.apache.cassandra.locator.DynamicEndpointSnitch
	Partitioner: org.apache.cassandra.dht.Murmur3Partitioner
	Schema versions:
		a34bf9e6-9139-3696-b85e-3fdd47aca08f: [172.17.0.2, 172.17.0.3]

		59adb24e-f3cd-3e02-97f0-5b395827453f: [172.17.0.5]
{code}

The node 172.17.0.5 is the one in 3.0.10. It has same keyspaces as other nodes.

When I use my application (protocol version to 3), it try to contact the 3.0.10 node and I get this error:

{code}
Cassandra timeout during write query at consistency LOCAL_ONE (1 replica were required but only 0 acknowledged the write)
{code}


For all my test, the seeder is 172.17.0.2 and replication factor is 1.

What is wrong in my migration ? Or is it a bug in the migration procedure ?
Why the blank node in 3.0.10 doesn't receive keyspaces description ?

Thanks.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)