GG-20702 Fix invalid partition clearing.

gridgain · Jul 3, 2019 · cafab0d · cafab0d
1 parent afdc96e
commit cafab0d
Show file tree

Hide file tree

Showing 5 changed files with 47 additions and 3 deletions.
diff --git a/...ache/ignite/internal/processors/cache/distributed/dht/topology/GridDhtLocalPartition.java b/...ache/ignite/internal/processors/cache/distributed/dht/topology/GridDhtLocalPartition.java
@@ -606,6 +606,8 @@ private boolean casState(long state, GridDhtPartitionState toState) {
      */
     public boolean own() {
         while (true) {
+            assert !clear : "Could not own clearing partition " + this;
+
             long state = this.state.get();
 
             GridDhtPartitionState partState = getPartState(state);

diff --git a/...nite/internal/processors/cache/distributed/dht/topology/GridDhtPartitionTopologyImpl.java b/...nite/internal/processors/cache/distributed/dht/topology/GridDhtPartitionTopologyImpl.java
@@ -74,6 +74,7 @@
 
 import static org.apache.ignite.events.EventType.EVT_CACHE_REBALANCE_PART_DATA_LOST;
 import static org.apache.ignite.internal.events.DiscoveryCustomEvent.EVT_DISCOVERY_CUSTOM_EVT;
+import static org.apache.ignite.internal.processors.cache.distributed.dht.preloader.ExchangeType.ALL;
 import static org.apache.ignite.internal.processors.cache.distributed.dht.topology.GridDhtPartitionState.EVICTED;
 import static org.apache.ignite.internal.processors.cache.distributed.dht.topology.GridDhtPartitionState.LOST;
 import static org.apache.ignite.internal.processors.cache.distributed.dht.topology.GridDhtPartitionState.MOVING;
@@ -763,7 +764,8 @@ private boolean partitionLocalNode(int p, AffinityTopologyVersion topVer) {
 
                 long updateSeq = this.updateSeq.incrementAndGet();
 
-                if (!ctx.localNode().isClient()) {
+                // Skip partition updates in case of not real exchange.
+                if (!ctx.localNode().isClient() && exchFut.exchangeType() == ALL) {
                     for (int p = 0; p < partitions; p++) {
                         GridDhtLocalPartition locPart = localPartition0(p, topVer, false, true);
 

diff --git a/...che/ignite/internal/processors/cache/distributed/dht/topology/PartitionsEvictManager.java b/...che/ignite/internal/processors/cache/distributed/dht/topology/PartitionsEvictManager.java
@@ -368,7 +368,7 @@ private void showProgress() {
                     ", grpId=" + grp.groupId() +
                     ", remainingPartsToEvict=" + (totalTasks.get() - taskInProgress) +
                     ", partsEvictInProgress=" + taskInProgress +
-                    ", totalParts= " + grp.topology().localPartitions().size() + "]");
+                    ", totalParts=" + grp.topology().localPartitions().size() + "]");
         }
     }
 

diff --git a/.../internal/processors/cache/persistence/db/IgnitePdsCacheWalDisabledOnRebalancingTest.java b/.../internal/processors/cache/persistence/db/IgnitePdsCacheWalDisabledOnRebalancingTest.java
@@ -345,6 +345,43 @@ public void testRebalancedPartitionsOwningWithConcurrentAffinityChange() throws
         verifyCache(ig1.cache(CACHE3_NAME), GENERATING_FUNC);
     }
 
+    /**
+     * Scenario: when rebalanced MOVING partitions are owning by checkpointer,
+     * concurrent no-op exchange should not trigger partition clearing.
+     *
+     * @throws Exception If failed.
+     */
+    @Test
+    public void testRebalancedPartitionsOwningWithAffinitySwitch() throws Exception {
+        Ignite ig0 = startGridsMultiThreaded(4);
+        fillCache(ig0.dataStreamer(CACHE3_NAME), CACHE_SIZE, GENERATING_FUNC);
+
+        // Stop idx=2 to prepare for baseline topology change later.
+        stopGrid(2);
+
+        // Stop idx=1 and cleanup LFS to trigger full rebalancing after it restart.
+        String ig1Name = "node01-" + grid(1).localNode().consistentId();
+        stopGrid(1);
+        cleanPersistenceFiles(ig1Name);
+
+        // Blocking fileIO and blockMessagePredicate to block checkpointer and rebalancing for node idx=1.
+        useBlockingFileIO = true;
+
+        // Enable blocking checkpointer on node idx=1 (see BlockingCheckpointFileIOFactory).
+        fileIoBlockingSemaphore.drainPermits();
+
+        // Wait for rebalance (all partitions will be in MOVING state until cp is finished).
+        startGrid(1).cachex(CACHE3_NAME).context().group().preloader().rebalanceFuture().get();
+
+        startGrid("client");
+
+        fileIoBlockingSemaphore.release(Integer.MAX_VALUE);
+
+        awaitPartitionMapExchange();
+
+        assertPartitionsSame(idleVerify(grid(0), CACHE3_NAME));
+    }
+
     /** FileIOFactory implementation that enables blocking of writes to disk so checkpoint can be blocked. */
     private static class BlockingCheckpointFileIOFactory implements FileIOFactory {
         /** Serial version uid. */

diff --git a/...e/src/test/java/org/apache/ignite/testframework/junits/common/GridCommonAbstractTest.java b/...e/src/test/java/org/apache/ignite/testframework/junits/common/GridCommonAbstractTest.java
@@ -872,6 +872,8 @@ protected void printPartitionState(String cacheName, int firstParts) {
 
             sb.append("nodeId=")
                 .append(k.context().localNodeId())
+                .append(" consistentId=")
+                .append(k.localNode().consistentId())
                 .append(" isDone=")
                 .append(syncFut.isDone())
                 .append("\n");
@@ -952,7 +954,8 @@ protected void printPartitionState(String cacheName, int firstParts) {
                         .append(part == null ? "NA" : part.dataStore().partUpdateCounter())
                         .append(" fullSize=")
                         .append(part == null ? "NA" : part.fullSize())
-                        .append(" state=").append(part.state());
+                        .append(" state=").append(part.state())
+                        .append(" reservations=").append(part.reservations());
                 }
                 else
                     sb.append(p).append(" is null");