Skip to content

HDDS-11128. Wait for replication to settle instead of a transient replica count #11010

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
smengcl wants to merge 3 commits into
base: master
Choose a base branch
from
+19 −2
Draft

HDDS-11128. Wait for replication to settle instead of a transient replica count #11010

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f712682
HDDS-11128. Increase timeout for decommission-triggered replica-count…
smengcl Aug 13, 2026
b8e518e
HDDS-11128. Reduce replica-count wait to 60s and disambiguate waitFor…
smengcl Aug 13, 2026
77a70a7
HDDS-11128. Wait for replication to settle instead of a transient rep…
smengcl Aug 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions ...test-recon/src/test/java/org/apache/hadoop/ozone/recon/TestReconAndAdminContainerCLI.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -272,7 +272,7 @@ void testNodesInDecommissionOrMaintenance(
// a new replica-copy is made to another node.
// For maintenance, there is no replica-copy in this case.
if (!isMaintenance) {
OzoneTestHelper.waitForReplicaCount(containerIdR3, 4, cluster);
OzoneTestHelper.waitForStableReplicaCount(containerIdR3, 4, cluster);
}

compareRMReportToReconResponse(underReplicatedState);
Expand All @@ -299,7 +299,7 @@ void testNodesInDecommissionOrMaintenance(
// There will be a replica copy for both maintenance and decommission.
// maintenance 3 -> 4, decommission 4 -> 5.
int expectedReplicaNum = isMaintenance ? 4 : 5;
OzoneTestHelper.waitForReplicaCount(containerIdR3, expectedReplicaNum, cluster);
OzoneTestHelper.waitForStableReplicaCount(containerIdR3, expectedReplicaNum, cluster);

compareRMReportToReconResponse(underReplicatedState);
compareRMReportToReconResponse(overReplicatedState);
Expand Down
17 changes: 17 additions & 0 deletions ...one/integration-test/src/test/java/org/apache/hadoop/ozone/container/OzoneTestHelper.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@
import org.apache.hadoop.hdds.scm.container.ContainerManager;
import org.apache.hadoop.hdds.scm.container.ContainerNotFoundException;
import org.apache.hadoop.hdds.scm.container.ContainerReplica;
import org.apache.hadoop.hdds.scm.container.replication.ReplicationManager;
import org.apache.hadoop.hdds.scm.events.SCMEvents;
import org.apache.hadoop.hdds.scm.pipeline.Pipeline;
import org.apache.hadoop.hdds.scm.pipeline.PipelineNotFoundException;
Expand Down Expand Up @@ -465,6 +466,22 @@ public static void waitForReplicaCount(long containerID, int count,
200, 30000);
}

/**
* Like {@link #waitForReplicaCount(long, int, MiniOzoneCluster)}, but only returns once
* replication has quiesced: no pending add or delete ops remain for the container, so the count
* has settled at {@code count} instead of being observed transiently while SCM is still adding or
* removing replicas. This removes the race where the count passes through {@code count} between
* polls under a loaded runner (HDDS-10582, HDDS-11128).
*/
public static void waitForStableReplicaCount(long containerID, int count, MiniOzoneCluster cluster)
throws TimeoutException, InterruptedException {
ReplicationManager replicationManager = cluster.getStorageContainerManager().getReplicationManager();
ContainerID cid = ContainerID.valueOf(containerID);
GenericTestUtils.waitFor(() ->
replicationManager.getPendingReplicationOps(cid).isEmpty() && countReplicas(containerID, cluster) == count,
200, 30000);
}

/**
* Wait until SCM reports exactly {@code count} replicas for the container and every replica is in {@code state}.
* Unlike {@link #waitForContainerStateInSCM}, which checks the container's aggregate state (it flips as soon as the
Expand Down
Loading