[
https://issues.apache.org/jira/browse/CASSANDRA-17883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17603178#comment-17603178
]
Jacek Lewandowski commented on CASSANDRA-17883:
-----------------------------------------------
We used this test to reproduce
{code:java}
try (Cluster cluster = init(Cluster.build(2).withConfig(cfg ->
cfg.with(Feature.GOSSIP, Feature.NETWORK)).start()))
{
cluster.schemaChange("CREATE TABLE " + KEYSPACE + ".tbl (pk INT
PRIMARY KEY, v TEXT)");
assertTrue(cluster.get(1).callOnInstance(() ->
Schema.instance.getTableMetadata(KEYSPACE, "tbl") != null));
assertTrue(cluster.get(2).callOnInstance(() ->
Schema.instance.getTableMetadata(KEYSPACE, "tbl") != null));
UUID hostId2 = cluster.get(2).callOnInstance(() ->
Gossiper.instance.getHostId(FBUtilities.getBroadcastAddressAndPort()));
cluster.get(2).shutdown().get();
Awaitility.await()
.atMost(ofSeconds(30))
.until(() -> cluster
.get(1)
.callOnInstance(() -> Gossiper.instance
.getLiveMembers()
.stream()
.allMatch(addr ->
addr.equals(FBUtilities.getBroadcastAddressAndPort()))));
int result = cluster.get(1).nodetool("removenode",
hostId2.toString());
Assertions.assertThat(result).isZero();
Awaitility.await()
.atMost(ofSeconds(30))
.until(() -> cluster.get(1).callOnInstance(() ->
Gossiper.instance.getUnreachableMembers().isEmpty()));
cluster.get(2).startup();
Awaitility.await()
.atMost(ofSeconds(30))
.until(() -> cluster
.get(1)
.callOnInstance(() -> Gossiper.instance
.getLiveMembers()
.size() == 2));
cluster.filters().verbs(Verb.SCHEMA_PUSH_REQ.id).drop();
cluster.coordinator(2).execute("CREATE TABLE " + KEYSPACE + ".tbl2
(pk INT PRIMARY KEY, v TEXT)", ConsistencyLevel.ONE);
Thread.sleep(30000);
cluster.filters().reset();
Awaitility.waitAtMost(ofMinutes(3))
.pollDelay(ofSeconds(1))
.until(() -> cluster.get(1).callOnInstance(() ->
Schema.instance.getTableMetadata(KEYSPACE, "tbl2") != null));
{code}
> Node will not get schema updates if it replaces a removed node with the same
> address
> ------------------------------------------------------------------------------------
>
> Key: CASSANDRA-17883
> URL: https://issues.apache.org/jira/browse/CASSANDRA-17883
> Project: Cassandra
> Issue Type: Bug
> Components: Legacy/Distributed Metadata
> Reporter: Jacek Lewandowski
> Priority: Normal
> Fix For: 3.0.x, 3.11.x, 4.0.x, 4.x
>
>
> If a node is removed and then a node with the same address is added, it is
> ignored by migration coordinator - that is, when that node fails to push the
> schema change, the other nodes will not be able pull it later during periodic
> update.
> Such situation will last until the other nodes gets restarted.
> This is caused by permanently adding the address of the node being removed to
> the ignore list in migration coordinator.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
