huizhilu commented on a change in pull request #1803:
URL: https://github.com/apache/helix/pull/1803#discussion_r659118244
##########
File path:
helix-core/src/main/java/org/apache/helix/api/status/ClusterManagementMode.java
##########
@@ -67,4 +69,26 @@ public Status getStatus() {
public Type getMode() {
return mode;
}
+
+ public boolean isFullyInNormalMode() {
Review comment:
What I'd like to have is:
1. What is the desired/target mode?
2. What is the status, IN PROGRESS or COMPLETED?
`isFullyInNormalMode` means the cluster is stable in the normal mode. For
the situation: recovering from pause, it should be IN_PROGRESS to NORMAL. It's
already set to NORMAL mode, so we want to mark it NORMAL, IN PROGRESS.
##########
File path:
helix-core/src/main/java/org/apache/helix/controller/stages/ManagementMessageDispatchStage.java
##########
@@ -0,0 +1,62 @@
+package org.apache.helix.controller.stages;
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+import java.util.List;
+
+import org.apache.helix.HelixManager;
+import org.apache.helix.api.status.ClusterManagementMode;
+import org.apache.helix.controller.LogUtil;
+import
org.apache.helix.controller.dataproviders.ManagementControllerDataProvider;
+import org.apache.helix.model.Message;
+import org.apache.helix.util.RebalanceUtil;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+/**
+ * Dispatches participant status change and pending state transition
cancellation messages
+ * for the management pipeline.
+ */
+public class ManagementMessageDispatchStage extends MessageDispatchStage {
+ private static final Logger LOG =
LoggerFactory.getLogger(ManagementMessageDispatchStage.class);
+
+ @Override
+ public void process(ClusterEvent event) throws Exception {
+ MessageOutput messageOutput =
+ event.getAttribute(AttributeName.MESSAGES_ALL.name());
+ processEvent(event, messageOutput);
+
+ // Send participant status change messages.
+ HelixManager manager =
event.getAttribute(AttributeName.helixmanager.name());
+ List<Message> messagesSent =
+ super.sendMessages(manager.getHelixDataAccessor(),
messageOutput.getStatusChangeMessages());
+ ManagementControllerDataProvider cache =
+ event.getAttribute(AttributeName.ControllerDataProvider.name());
+ cache.cacheMessages(messagesSent);
+
+ // Can exit management mode pipeline after fully in normal mode
+ ClusterManagementMode managementMode =
event.getAttribute(AttributeName.CLUSTER_STATUS.name());
+ if (managementMode.isFullyInNormalMode()) {
Review comment:
We want to finish running the pipeline. And at the final step, decide
whether if it can exit.
The pros:
* we don't need to throw an exception to terminate the pipeline.
* It makes sense to determine if it should switch at the final step after
finishing all the previous checks.
##########
File path:
helix-core/src/main/java/org/apache/helix/controller/stages/ManagementModeStage.java
##########
@@ -66,6 +68,14 @@ public void process(ClusterEvent event) throws Exception {
HelixDataAccessor accessor = manager.getHelixDataAccessor();
ManagementControllerDataProvider cache =
event.getAttribute(AttributeName.ControllerDataProvider.name());
+ CurrentStateOutput currentStateOutput =
+ event.getAttribute(AttributeName.CURRENT_STATE.name());
+ final Map<String, Resource> resourceMap =
+ event.getAttribute(AttributeName.RESOURCES_TO_REBALANCE.name());
+
+ final BestPossibleStateOutput bestPossibleStateOutput =
Review comment:
Replied in the above thread. I don't think it's that far. It's used in
the next stage. This stage prepares required output for the next stage. It
makes sense, also aligned with the default/task pipelines.
##########
File path:
helix-core/src/main/java/org/apache/helix/controller/stages/ManagementMessageGenerationPhase.java
##########
@@ -0,0 +1,118 @@
+package org.apache.helix.controller.stages;
+
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+import org.apache.helix.HelixManager;
+import org.apache.helix.api.status.ClusterManagementMode;
+import org.apache.helix.controller.LogUtil;
+import
org.apache.helix.controller.dataproviders.ManagementControllerDataProvider;
+import org.apache.helix.controller.pipeline.StageException;
+import org.apache.helix.model.LiveInstance;
+import org.apache.helix.model.LiveInstance.LiveInstanceStatus;
+import org.apache.helix.model.Message;
+import org.apache.helix.model.PauseSignal;
+import org.apache.helix.util.MessageUtil;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+/**
+ * Generates participant status change (freeze/unfreeze) and pending state
transition cancellation
+ * messages for management mode pipeline.
+ */
+public class ManagementMessageGenerationPhase extends MessageGenerationPhase {
+ private static final Logger LOG =
LoggerFactory.getLogger(ManagementMessageGenerationPhase.class);
+
+ @Override
+ public void process(ClusterEvent event) throws Exception {
+ _eventId = event.getEventId();
+ String clusterName = event.getClusterName();
+ HelixManager manager =
event.getAttribute(AttributeName.helixmanager.name());
+ ClusterManagementMode managementMode =
event.getAttribute(AttributeName.CLUSTER_STATUS.name());
+ ManagementControllerDataProvider cache =
+ event.getAttribute(AttributeName.ControllerDataProvider.name());
+ if (manager == null || managementMode == null || cache == null) {
+ throw new StageException("Missing attributes in event: " + event
+ + ". Requires HelixManager|ClusterStatus|DataCache");
+ }
+
+ PauseSignal pauseSignal = cache.getPauseSignal();
+ if (cache.getClusterConfig().isStateTransitionCancelEnabled()
+ && pauseSignal != null && pauseSignal.getCancelPendingST()) {
+ // Generate ST cancellation messages.
+ LogUtil.logInfo(LOG, _eventId,
+ "Generating ST cancellation messages for cluster " + clusterName);
+ super.process(event);
Review comment:
I'd like to make the stages clearly do what it is supposed to do:
* ManagementModeStage: check status, prepare for message generation.
* ManagementMessageGenerationPhase: only create messages, based on the
results of the previous stages.
I don't want to mix the responsibilities. So I think putting the logic: CS
-> best possible in management mode stage is appropriate as preparation. It
also aligns with the logic as the default pipeline: messages are generated
purely based on the previous results.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
