Fixed the test timeout by setting LogAppenderWaitTimeMin for Ratis

hemantk-12 · Jan 22, 2024 · a059ae6 · a059ae6
1 parent 18fc1ec
commit a059ae6
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 11 deletions.
diff --git a/...b-checkpoint-differ/src/main/java/org/apache/ozone/rocksdiff/RocksDBCheckpointDiffer.java b/...b-checkpoint-differ/src/main/java/org/apache/ozone/rocksdiff/RocksDBCheckpointDiffer.java
@@ -1453,8 +1453,13 @@ public void incrementTarballRequestCount() {
     tarballRequestCount.incrementAndGet();
   }
 
-  public void decrementTarballRequestCount() {
-    tarballRequestCount.decrementAndGet();
+  public void decrementTarballRequestCountAndNotify() {
+    // Synchronized block is used to ensure that lock is on the same instance notifyAll is being called.
+    synchronized (this) {
+      tarballRequestCount.decrementAndGet();
+      // Notify compaction threads to continue.
+      notifyAll();
+    }
   }
 
   public boolean shouldRun() {

diff --git a/...gration-test/src/test/java/org/apache/hadoop/ozone/om/TestSnapshotBackgroundServices.java b/...gration-test/src/test/java/org/apache/hadoop/ozone/om/TestSnapshotBackgroundServices.java
@@ -39,6 +39,7 @@
 import org.apache.hadoop.ozone.om.helpers.BucketLayout;
 import org.apache.hadoop.ozone.om.helpers.OmKeyInfo;
 import org.apache.hadoop.ozone.om.helpers.SnapshotInfo;
+import org.apache.hadoop.ozone.om.ratis.OzoneManagerRatisServerConfig;
 import org.apache.hadoop.ozone.om.snapshot.ReferenceCounted;
 import org.apache.hadoop.ozone.om.snapshot.SnapshotCache;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffReportOzone;
@@ -114,11 +115,12 @@ public void init(TestInfo testInfo) throws Exception {
     String clusterId = UUID.randomUUID().toString();
     String scmId = UUID.randomUUID().toString();
     String omServiceId = "om-service-test1";
+    OzoneManagerRatisServerConfig omRatisConf = conf.getObject(OzoneManagerRatisServerConfig.class);
+    omRatisConf.setLogAppenderWaitTimeMin(10);
+    conf.setFromObject(omRatisConf);
     conf.setInt(OMConfigKeys.OZONE_OM_RATIS_LOG_PURGE_GAP, LOG_PURGE_GAP);
-    conf.setStorageSize(OMConfigKeys.OZONE_OM_RATIS_SEGMENT_SIZE_KEY, 16,
-        StorageUnit.KB);
-    conf.setStorageSize(OMConfigKeys.
-        OZONE_OM_RATIS_SEGMENT_PREALLOCATED_SIZE_KEY, 16, StorageUnit.KB);
+    conf.setStorageSize(OMConfigKeys.OZONE_OM_RATIS_SEGMENT_SIZE_KEY, 16, StorageUnit.KB);
+    conf.setStorageSize(OMConfigKeys.OZONE_OM_RATIS_SEGMENT_PREALLOCATED_SIZE_KEY, 16, StorageUnit.KB);
     if ("testSSTFilteringBackgroundService".equals(testInfo.getDisplayName())) {
       conf.setTimeDuration(OZONE_SNAPSHOT_SST_FILTERING_SERVICE_INTERVAL, 1,
           TimeUnit.SECONDS);

diff --git a/...p-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OMDBCheckpointServlet.java b/...p-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OMDBCheckpointServlet.java
@@ -241,9 +241,9 @@ public DBCheckpoint getCheckpoint(Path tmpdir, boolean flush)
     long startTime = System.currentTimeMillis();
     long pauseCounter = PAUSE_COUNTER.incrementAndGet();
 
-    // Pause compactions, Copy/link files and get checkpoint.
     try {
       LOG.info("Compaction pausing {} started.", pauseCounter);
+      // Pause compactions, Copy/link files and get checkpoint.
       differ.incrementTarballRequestCount();
       FileUtils.copyDirectory(compactionLogDir.getOriginalDir(),
           compactionLogDir.getTmpDir());
@@ -252,10 +252,7 @@ public DBCheckpoint getCheckpoint(Path tmpdir, boolean flush)
       checkpoint = getDbStore().getCheckpoint(flush);
     } finally {
       // Unpause the compaction threads.
-      synchronized (getDbStore().getRocksDBCheckpointDiffer()) {
-        differ.decrementTarballRequestCount();
-        differ.notifyAll();
-      }
+      differ.decrementTarballRequestCountAndNotify();
       long elapsedTime = System.currentTimeMillis() - startTime;
       LOG.info("Compaction pausing {} ended. Elapsed ms: {}", pauseCounter, elapsedTime);
     }