Fix file reading in ccr restore service (#38117)
Currently we use the raw byte array length when calling the IndexInput read call to determine how many bytes we want to read. However, due to how BigArrays works, the array length might be longer than the reference length. This commit fixes the issue and uses the BytesRef length when calling read. Additionally, it expands the index follow test to index many more documents. These documents should potentially lead to large enough segment files to trigger scenarios where this fix matters.
This commit is contained in:
parent
c67a9663af
commit
291c4e7a0c
|
@ -235,8 +235,7 @@ public class CcrRestoreSourceService extends AbstractLifecycleComponent implemen
|
|||
BytesRefIterator refIterator = reference.iterator();
|
||||
BytesRef ref;
|
||||
while ((ref = refIterator.next()) != null) {
|
||||
byte[] refBytes = ref.bytes;
|
||||
indexInput.readBytes(refBytes, 0, refBytes.length);
|
||||
indexInput.readBytes(ref.bytes, ref.offset, ref.length);
|
||||
}
|
||||
|
||||
long offsetAfterRead = indexInput.getFilePointer();
|
||||
|
|
|
@ -70,7 +70,6 @@ import org.elasticsearch.transport.TransportService;
|
|||
import org.elasticsearch.transport.nio.MockNioTransportPlugin;
|
||||
import org.elasticsearch.xpack.ccr.CcrSettings;
|
||||
import org.elasticsearch.xpack.ccr.LocalStateCcr;
|
||||
import org.elasticsearch.xpack.ccr.index.engine.FollowingEngine;
|
||||
import org.elasticsearch.xpack.core.XPackSettings;
|
||||
import org.elasticsearch.xpack.core.ccr.AutoFollowMetadata;
|
||||
import org.elasticsearch.xpack.core.ccr.ShardFollowNodeTaskStatus;
|
||||
|
@ -551,27 +550,6 @@ public abstract class CcrIntegTestCase extends ESTestCase {
|
|||
});
|
||||
}
|
||||
|
||||
protected void assertTotalNumberOfOptimizedIndexing(Index followerIndex, int numberOfShards, long expectedTotal) throws Exception {
|
||||
assertBusy(() -> {
|
||||
long[] numOfOptimizedOps = new long[numberOfShards];
|
||||
for (int shardId = 0; shardId < numberOfShards; shardId++) {
|
||||
for (String node : getFollowerCluster().nodesInclude(followerIndex.getName())) {
|
||||
IndicesService indicesService = getFollowerCluster().getInstance(IndicesService.class, node);
|
||||
IndexShard shard = indicesService.getShardOrNull(new ShardId(followerIndex, shardId));
|
||||
if (shard != null && shard.routingEntry().primary()) {
|
||||
try {
|
||||
FollowingEngine engine = ((FollowingEngine) IndexShardTestCase.getEngine(shard));
|
||||
numOfOptimizedOps[shardId] = engine.getNumberOfOptimizedIndexing();
|
||||
} catch (AlreadyClosedException e) {
|
||||
throw new AssertionError(e); // causes assertBusy to retry
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
assertThat(Arrays.stream(numOfOptimizedOps).sum(), equalTo(expectedTotal));
|
||||
});
|
||||
}
|
||||
|
||||
static void removeCCRRelatedMetadataFromClusterState(ClusterService clusterService) throws Exception {
|
||||
CountDownLatch latch = new CountDownLatch(1);
|
||||
clusterService.submitStateUpdateTask("remove-ccr-related-metadata", new ClusterStateUpdateTask() {
|
||||
|
|
|
@ -26,6 +26,7 @@ import org.elasticsearch.action.admin.indices.settings.put.UpdateSettingsRequest
|
|||
import org.elasticsearch.action.admin.indices.stats.ShardStats;
|
||||
import org.elasticsearch.action.bulk.BulkProcessor;
|
||||
import org.elasticsearch.action.bulk.BulkRequest;
|
||||
import org.elasticsearch.action.bulk.BulkRequestBuilder;
|
||||
import org.elasticsearch.action.bulk.BulkResponse;
|
||||
import org.elasticsearch.action.get.GetResponse;
|
||||
import org.elasticsearch.action.index.IndexRequest;
|
||||
|
@ -101,9 +102,30 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
assertAcked(leaderClient().admin().indices().prepareCreate("index1").setSource(leaderIndexSettings, XContentType.JSON));
|
||||
ensureLeaderYellow("index1");
|
||||
|
||||
final int firstBatchNumDocs = randomIntBetween(2, 64);
|
||||
final int firstBatchNumDocs;
|
||||
// Sometimes we want to index a lot of documents to ensure that the recovery works with larger files
|
||||
if (rarely()) {
|
||||
firstBatchNumDocs = randomIntBetween(1800, 2000);
|
||||
} else {
|
||||
firstBatchNumDocs = randomIntBetween(10, 64);
|
||||
}
|
||||
final int flushPoint = (int) (firstBatchNumDocs * 0.75);
|
||||
|
||||
logger.info("Indexing [{}] docs as first batch", firstBatchNumDocs);
|
||||
for (int i = 0; i < firstBatchNumDocs; i++) {
|
||||
BulkRequestBuilder bulkRequestBuilder = leaderClient().prepareBulk();
|
||||
for (int i = 0; i < flushPoint; i++) {
|
||||
final String source = String.format(Locale.ROOT, "{\"f\":%d}", i);
|
||||
IndexRequest indexRequest = new IndexRequest("index1", "doc", Integer.toString(i))
|
||||
.source(source, XContentType.JSON)
|
||||
.timeout(TimeValue.timeValueSeconds(1));
|
||||
bulkRequestBuilder.add(indexRequest);
|
||||
}
|
||||
bulkRequestBuilder.get();
|
||||
|
||||
leaderClient().admin().indices().prepareFlush("index1").setWaitIfOngoing(true).get();
|
||||
|
||||
// Index some docs after the flush that might be recovered in the normal index following operations
|
||||
for (int i = flushPoint; i < firstBatchNumDocs; i++) {
|
||||
final String source = String.format(Locale.ROOT, "{\"f\":%d}", i);
|
||||
leaderClient().prepareIndex("index1", "doc", Integer.toString(i)).setSource(source, XContentType.JSON).get();
|
||||
}
|
||||
|
@ -147,7 +169,7 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
for (int i = 0; i < firstBatchNumDocs; i++) {
|
||||
assertBusy(assertExpectedDocumentRunnable(i));
|
||||
}
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), numberOfPrimaryShards, firstBatchNumDocs);
|
||||
|
||||
pauseFollow("index2");
|
||||
followerClient().execute(ResumeFollowAction.INSTANCE, followRequest.getFollowRequest()).get();
|
||||
final int secondBatchNumDocs = randomIntBetween(2, 64);
|
||||
|
@ -172,8 +194,6 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
for (int i = firstBatchNumDocs; i < firstBatchNumDocs + secondBatchNumDocs; i++) {
|
||||
assertBusy(assertExpectedDocumentRunnable(i));
|
||||
}
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), numberOfPrimaryShards,
|
||||
firstBatchNumDocs + secondBatchNumDocs);
|
||||
pauseFollow("index2");
|
||||
assertMaxSeqNoOfUpdatesIsTransferred(resolveLeaderIndex("index1"), resolveFollowerIndex("index2"), numberOfPrimaryShards);
|
||||
}
|
||||
|
@ -287,7 +307,6 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
for (int i = 0; i < firstBatchNumDocs; i++) {
|
||||
assertBusy(assertExpectedDocumentRunnable(i));
|
||||
}
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), numberOfPrimaryShards, firstBatchNumDocs);
|
||||
pauseFollow("index2");
|
||||
}
|
||||
|
||||
|
@ -432,8 +451,6 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
assertIndexFullyReplicatedToFollower("index1", "index2");
|
||||
pauseFollow("index2");
|
||||
leaderClient().admin().indices().prepareRefresh("index1").get();
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), numberOfShards,
|
||||
leaderClient().prepareSearch("index1").get().getHits().getTotalHits().value);
|
||||
assertMaxSeqNoOfUpdatesIsTransferred(resolveLeaderIndex("index1"), resolveFollowerIndex("index2"), numberOfShards);
|
||||
}
|
||||
|
||||
|
@ -475,7 +492,6 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
}
|
||||
pauseFollow("index2");
|
||||
assertMaxSeqNoOfUpdatesIsTransferred(resolveLeaderIndex("index1"), resolveFollowerIndex("index2"), 1);
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), 1, numDocs);
|
||||
}
|
||||
|
||||
public void testUnfollowNonExistingIndex() {
|
||||
|
@ -538,7 +554,6 @@ public class IndexFollowingIT extends CcrIntegTestCase {
|
|||
}
|
||||
pauseFollow("index2");
|
||||
assertMaxSeqNoOfUpdatesIsTransferred(resolveLeaderIndex("index1"), resolveFollowerIndex("index2"), 1);
|
||||
assertTotalNumberOfOptimizedIndexing(resolveFollowerIndex("index2"), 1, numDocs);
|
||||
}
|
||||
|
||||
public void testAttemptToChangeCcrFollowingIndexSetting() throws Exception {
|
||||
|
|
Loading…
Reference in New Issue