rpuch commented on code in PR #2473:
URL: https://github.com/apache/ignite-3/pull/2473#discussion_r1312710088
##########
modules/runner/src/integrationTest/java/org/apache/ignite/internal/raftsnapshot/ItTableRaftSnapshotsTest.java:
##########
@@ -792,6 +816,116 @@ void
testChangeLeaderDuringSnapshotInstallationToLeaderWithEnoughLog() throws Ex
assertThat(rows, is(List.of(new IgniteBiTuple<>(1, "one"))));
}
+ /**
+ * The replication mechanism must not replicate commands for which schemas
are not yet available on the node
+ * to which replication happens (in Raft, it means that followers/learners
cannot receive commands that they
+ * cannot execute without waiting for schemas). This method tests that
snapshots bringing such commands are
+ * rejected, and that, when metadata catches up, the snapshot gets
successfully installed.
+ */
+ @Test
+ void laggingSchemasOnFollowerPreventSnapshotInstallation() throws
Exception {
+ cluster.startAndInit(3);
+
+ createTestTableWith3Replicas(DEFAULT_STORAGE_ENGINE);
+
+ // Prepare the scene: force node 0 to be a leader, and node 2 to be a
follower.
+
+ final int leaderIndex = 0;
+ final int followerIndex = 2;
+
+ transferLeadershipOnSolePartitionTo(leaderIndex);
+ cluster.transferLeadershipTo(leaderIndex, MetastorageGroupId.INSTANCE);
+
+ // Block AppendEntries from being accepted on the follower so that the
leader will have to use a snapshot.
+ blockIncomingAppendEntriesAt(followerIndex);
Review Comment:
The problem is that if node C's watches are inhibited it will not see the
assignments update, so the rebalance will not happen.
Also, using regular mechanisms in such a way looks as a hack itself. It
seems that when we want to cripple the ability of a node to receive
AppendEntries we should do exactly this, it makes the test a little bit
understandable and less fragile.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
