nsivabalan commented on code in PR #13316:
URL: https://github.com/apache/hudi/pull/13316#discussion_r2101431783


##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/transaction/TableServicePlanConflictDetection.java:
##########
@@ -0,0 +1,106 @@
+/*
+ * 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.
+ */
+
+package org.apache.hudi.client.transaction;
+
+import org.apache.hudi.avro.model.HoodieClusteringPlan;
+import org.apache.hudi.avro.model.HoodieCompactionPlan;
+import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.common.table.timeline.HoodieInstant;
+import org.apache.hudi.exception.HoodieIOException;
+
+import java.io.IOException;
+import java.util.stream.Stream;
+
+/**
+ * Handles conflict detection for the plans of various table services.
+ * Cleaning and archiving are not included in this class as they do not 
require conflict detection.
+ */
+public class TableServicePlanConflictDetection {
+  private final ConflictResolutionStrategy conflictResolutionStrategy;
+  private final HoodieTableMetaClient metaClient;
+  private final String lastKnownCompletionTime;
+
+  /**
+   * Constructors a new instance to detect conflicts for a table service plan.
+   * Note that this should be used within a transaction to prevent race 
conditions. The meta client used must be constructed or reloaded within the 
transaction.
+   * @param conflictResolutionStrategy The strategy to use for conflict 
resolution for the writer.
+   * @param metaClient The meta client for the table. This must be constructed 
or reloaded during the transaction.
+   * @param lastKnownCompletionTime The last completion time of an instant on 
the timeline when the planner started.
+   */
+  public TableServicePlanConflictDetection(ConflictResolutionStrategy 
conflictResolutionStrategy, HoodieTableMetaClient metaClient, String 
lastKnownCompletionTime) {
+    this.conflictResolutionStrategy = conflictResolutionStrategy;
+    this.metaClient = metaClient;
+    this.lastKnownCompletionTime = lastKnownCompletionTime;
+  }
+
+  /**
+   * Checks if there are any potential conflicts with actions that have 
happened since the planner start instant.
+   * The compaction requires a check for conflicts with any commits for 
ingesting new data since the compaction plan is expected to contain all the log 
files for the slice it is compacting.
+   * It also requires checks for new plans for clustering or potentially 
another compaction on the timeline to avoid multiple services operating on the 
same file groups.
+   * @param compactionPlan the requested plan
+   * @param requestedInstantTime the requested instant time that will be used 
when persisting the plan to the timeline
+   * @return true if there is a conflict with the requested plan, false 
otherwise
+   */
+  public boolean hasConflict(HoodieCompactionPlan compactionPlan, String 
requestedInstantTime) {
+    ConcurrentOperation compactionOperation = new 
ConcurrentOperation(compactionPlan, requestedInstantTime);
+    return hasConflict(compactionOperation);
+  }
+
+  /**
+   * Checks if there are any potential conflicts with actions that have 
happened since the planner start instant.
+   * The clustering requires a check for conflicts with any commits for 
ingesting new data since the clustering plan will also contain log files,
+   * and all the log files for the slice it is clustering must be in the plan.
+   * It also requires checks for new plans for compaction or potentially 
another clustering on the timeline to avoid multiple services operating on the 
same file groups.
+   * @param clusteringPlan the requested plan
+   * @param requestedInstantTime the requested instant time that will be used 
when persisting the plan to the timeline
+   * @return true if there is a conflict with the requested plan, false 
otherwise
+   */
+  public boolean hasConflict(HoodieClusteringPlan clusteringPlan, String 
requestedInstantTime) {
+    ConcurrentOperation clusteringOperation = new 
ConcurrentOperation(clusteringPlan, requestedInstantTime);
+    return hasConflict(clusteringOperation);
+  }
+
+  private boolean hasConflict(ConcurrentOperation plannedOperation) {
+    return generateCandidateInstants().anyMatch(instant -> {
+      try {
+        ConcurrentOperation otherOperation = new ConcurrentOperation(instant, 
metaClient);
+        return conflictResolutionStrategy.hasConflict(plannedOperation, 
otherOperation);

Review Comment:
   but here is my understanding. 
   
   we need to distinguish if its an inflight ingestion commit or completed. 
   
   a. completed ingestion commit. and compaction plan that just got generated 
is going through conflict resolution. 
   if there are overlapping file groups, we are failing here. bcoz, the cur 
plan generated would not have included log files produced by an inglith 
ingestion commit. 
   
   b. inflight ingestion commit. and compaction plan that just got generated is 
going through conflict resolution. 
   here we let the compaction plan go ahead. since compaction plan conflict 
resolution should be happening within a lock and once the resolution succeeds, 
compaction.requested will be added to the timeline. and hence the ingestion 
commit's completion time will definitely be > the compaction instant time. 
   
   in this case, lets also see how does conflict resolution of entire write 
will happen for ingestion commit (I am not talking about 
TableServicePlanConflictDetection). 
   - Guess we might need a fix here. currently(as per master), if the ingestion 
commits's start time is < compaction instant time, we fail the resolution step. 
if its  >, we let is pass. 
   we might need to make changes to this once we have this patch landed. 
essentially, ingestion commits can never conflict w/ a compaction commit during 
the final conflict resolution. 
   
   But I will let Tim also go over this and see if my understanding is right. 
   
   



-- 
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]

Reply via email to