umamaheswararao commented on a change in pull request #3127:
URL: https://github.com/apache/ozone/pull/3127#discussion_r814485024
##########
File path:
hadoop-ozone/client/src/test/java/org/apache/hadoop/ozone/client/TestOzoneECClient.java
##########
@@ -908,6 +911,107 @@ public void testPartialStripeWithPartialChunkRetry()
}
}
+ @Test
+ public void testDiscardPreAllocatedBlocksPreventRetryExceeds()
+ throws IOException {
+ close();
+ OzoneConfiguration con = new OzoneConfiguration();
+ int maxRetries = 3;
+ con.setStorageSize(OzoneConfigKeys.OZONE_SCM_BLOCK_SIZE,
+ 2, StorageUnit.KB);
+ con.setInt(OzoneConfigKeys.OZONE_CLIENT_MAX_EC_STRIPE_WRITE_RETRIES,
+ maxRetries);
+ MultiNodePipelineBlockAllocator blkAllocator =
+ new MultiNodePipelineBlockAllocator(con, dataBlocks + parityBlocks,
+ 15);
+ createNewClient(con, blkAllocator);
+
+ store.createVolume(volumeName);
+ OzoneVolume volume = store.getVolume(volumeName);
+ volume.createBucket(bucketName);
+ OzoneBucket bucket = volume.getBucket(bucketName);
+
+ int numStripesBeforeFailure = 1;
+ int numStripesAfterFailure = 1;
+ int numSTripesTotal = numStripesBeforeFailure + numStripesAfterFailure;
+ int numChunksToWriteAfterFailure = dataBlocks;
+ int numExpectedBlockGrps = 2;
+ // fail the DNs for parity blocks
+ int[] nodesIndexesToMarkFailure = {3, 4};
+
+ try (OzoneOutputStream out = bucket.createKey(keyName,
+ 1024 * dataBlocks * numStripesBeforeFailure
+ + numChunksToWriteAfterFailure,
+ new ECReplicationConfig(dataBlocks, parityBlocks,
+ ECReplicationConfig.EcCodec.RS,
+ chunkSize), new HashMap<>())) {
+ Assert.assertTrue(out.getOutputStream() instanceof ECKeyOutputStream);
+ ECKeyOutputStream kos = (ECKeyOutputStream) out.getOutputStream();
+ List<OmKeyLocationInfo> blockInfos = kos.getAllLocationInfoList();
+ Assert.assertEquals(1, blockInfos.size());
+
+ // Mock some pre-allocated blocks to the key
+ // should be > maxRetries
+ int numPreAllocatedBlocks = maxRetries + 1;
+ BlockID blockID = blockInfos.get(0).getBlockID();
+ Pipeline pipeline = blockInfos.get(0).getPipeline();
+ List<OmKeyLocationInfo> omKeyLocationInfos = new ArrayList<>();
+ for (int i = 0; i < numPreAllocatedBlocks; i++) {
+ BlockID nextBlockID = new BlockID(blockID.getContainerID(),
+ blockID.getLocalID() + i + 1);
+ omKeyLocationInfos.add(new OmKeyLocationInfo.Builder()
+ .setBlockID(nextBlockID)
+ .setPipeline(pipeline)
+ .build());
+ }
+ OmKeyLocationInfoGroup omKeyLocationInfoGroup =
+ new OmKeyLocationInfoGroup(0, omKeyLocationInfos);
+ kos.addPreallocateBlocks(omKeyLocationInfoGroup, 0);
+
+ for (int j = 0; j < numStripesBeforeFailure; j++) {
+ for (int i = 0; i < dataBlocks; i++) {
+ out.write(inputChunks[i]);
+ }
+ }
+
+ // Make the parity write fail to trigger retry
+ List<DatanodeDetails> failedDNs = new ArrayList<>();
+ List<HddsProtos.DatanodeDetailsProto> dns = allocator.getClusterDns();
+ for (int j = 0; j < nodesIndexesToMarkFailure.length; j++) {
+ failedDNs.add(DatanodeDetails
+ .getFromProtoBuf(dns.get(nodesIndexesToMarkFailure[j])));
+ }
+
+ // First let's set storage as bad
+ ((MockXceiverClientFactory) factoryStub).setFailedStorages(failedDNs);
+
+ for (int j = 0; j < numStripesAfterFailure; j++) {
+ for (int i = 0; i < dataBlocks; i++) {
+ out.write(inputChunks[i]);
+ }
+ }
+
+ // if we don't discard pre-allocated blocks,
Review comment:
comment can be like "if we don't discard pre-allocated blocks, retries
should exceed the maxRetries and write will fail." ?
##########
File path:
hadoop-ozone/client/src/test/java/org/apache/hadoop/ozone/client/TestOzoneECClient.java
##########
@@ -908,6 +911,107 @@ public void testPartialStripeWithPartialChunkRetry()
}
}
+ @Test
+ public void testDiscardPreAllocatedBlocksPreventRetryExceeds()
+ throws IOException {
+ close();
+ OzoneConfiguration con = new OzoneConfiguration();
+ int maxRetries = 3;
+ con.setStorageSize(OzoneConfigKeys.OZONE_SCM_BLOCK_SIZE,
+ 2, StorageUnit.KB);
+ con.setInt(OzoneConfigKeys.OZONE_CLIENT_MAX_EC_STRIPE_WRITE_RETRIES,
+ maxRetries);
+ MultiNodePipelineBlockAllocator blkAllocator =
+ new MultiNodePipelineBlockAllocator(con, dataBlocks + parityBlocks,
+ 15);
+ createNewClient(con, blkAllocator);
+
+ store.createVolume(volumeName);
+ OzoneVolume volume = store.getVolume(volumeName);
+ volume.createBucket(bucketName);
+ OzoneBucket bucket = volume.getBucket(bucketName);
+
+ int numStripesBeforeFailure = 1;
+ int numStripesAfterFailure = 1;
+ int numSTripesTotal = numStripesBeforeFailure + numStripesAfterFailure;
+ int numChunksToWriteAfterFailure = dataBlocks;
+ int numExpectedBlockGrps = 2;
+ // fail the DNs for parity blocks
+ int[] nodesIndexesToMarkFailure = {3, 4};
+
+ try (OzoneOutputStream out = bucket.createKey(keyName,
Review comment:
Below number seems key for getting the pre-allocated blocks. Can we
extract to a variable and provide one line comment with details like how many
pre-allocated blocks expected? But looks like we are adding preallocated blocks
explicitly. Then this number really matters?
##########
File path:
hadoop-ozone/client/src/main/java/org/apache/hadoop/ozone/client/io/ECKeyOutputStream.java
##########
@@ -87,6 +88,23 @@ public XceiverClientFactory getXceiverClientFactory() {
return blockOutputStreamEntryPool.getLocationInfoList();
}
+ @VisibleForTesting
Review comment:
Looks like this used only for tests?
If so, why can't we use getStreamEntries API and build the required info in
tests?
##########
File path:
hadoop-ozone/client/src/test/java/org/apache/hadoop/ozone/client/TestOzoneECClient.java
##########
@@ -908,6 +911,107 @@ public void testPartialStripeWithPartialChunkRetry()
}
}
+ @Test
+ public void testDiscardPreAllocatedBlocksPreventRetryExceeds()
+ throws IOException {
+ close();
+ OzoneConfiguration con = new OzoneConfiguration();
+ int maxRetries = 3;
+ con.setStorageSize(OzoneConfigKeys.OZONE_SCM_BLOCK_SIZE,
+ 2, StorageUnit.KB);
+ con.setInt(OzoneConfigKeys.OZONE_CLIENT_MAX_EC_STRIPE_WRITE_RETRIES,
+ maxRetries);
+ MultiNodePipelineBlockAllocator blkAllocator =
+ new MultiNodePipelineBlockAllocator(con, dataBlocks + parityBlocks,
+ 15);
+ createNewClient(con, blkAllocator);
+
+ store.createVolume(volumeName);
+ OzoneVolume volume = store.getVolume(volumeName);
+ volume.createBucket(bucketName);
+ OzoneBucket bucket = volume.getBucket(bucketName);
+
+ int numStripesBeforeFailure = 1;
+ int numStripesAfterFailure = 1;
+ int numSTripesTotal = numStripesBeforeFailure + numStripesAfterFailure;
+ int numChunksToWriteAfterFailure = dataBlocks;
+ int numExpectedBlockGrps = 2;
+ // fail the DNs for parity blocks
Review comment:
There is no reason to make only parity nodes fail right? we can fail any
nodes to trigger retry 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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
