[
https://issues.apache.org/jira/browse/IGNITE-21001?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mikhail Petrov updated IGNITE-21001:
------------------------------------
Description:
When a client requests information about cluster nodes, it sends the version of
the topology for which it received a list of cluster nodes the previous time.
And if the client's version of the topology matches the one currently on the
server, then the server does not send the list of cluster nodes back to the
client. Like the client knows everything itself.
But after cluster restart, the topology version may not change. But the nodes
will be different.
The reproducer:
{noformat}
@Test
public void testClusterGroup() throws Exception {
IgniteEx srv = startGrid(0);
try (IgniteClient client = startClient(0)) {
assertTrue(client.cluster().nodes().contains(srv.localNode()));
stopAllGrids();
srv = startGrid(0);
assertTrue(client.cluster().nodes().contains(srv.localNode()));
}
}
{noformat}
was:
When a client requests information about cluster nodes, it sends the version of
the topology for which it received a list of cluster nodes the previous time.
And if the client's version of the topology matches the one currently on the
server, then the server does not send the list of cluster nodes back to the
client. Like the client knows everything itself.
But after restarting the cluster, the topology version may not change. But the
nodes will be different.
The reproducer:
{noformat}
@Test
public void testClusterGroup() throws Exception {
IgniteEx srv = startGrid(0);
try (IgniteClient client = startClient(0)) {
assertTrue(client.cluster().nodes().contains(srv.localNode()));
stopAllGrids();
srv = startGrid(0);
assertTrue(client.cluster().nodes().contains(srv.localNode()));
}
}
{noformat}
> Thin client: cluster group returns cached nodes after cluster restart
> ---------------------------------------------------------------------
>
> Key: IGNITE-21001
> URL: https://issues.apache.org/jira/browse/IGNITE-21001
> Project: Ignite
> Issue Type: Bug
> Reporter: Nikita Amelchev
> Assignee: Mikhail Petrov
> Priority: Critical
> Labels: ise
> Fix For: 2.17
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> When a client requests information about cluster nodes, it sends the version
> of the topology for which it received a list of cluster nodes the previous
> time.
> And if the client's version of the topology matches the one currently on the
> server, then the server does not send the list of cluster nodes back to the
> client. Like the client knows everything itself.
> But after cluster restart, the topology version may not change. But the nodes
> will be different.
> The reproducer:
> {noformat}
> @Test
> public void testClusterGroup() throws Exception {
> IgniteEx srv = startGrid(0);
> try (IgniteClient client = startClient(0)) {
> assertTrue(client.cluster().nodes().contains(srv.localNode()));
> stopAllGrids();
> srv = startGrid(0);
> assertTrue(client.cluster().nodes().contains(srv.localNode()));
> }
> }
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
