honeymoose
/
OpenSearch
mirror of https://github.com/honeymoose/OpenSearch.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Make BackgroundIndexer more Efficient (#57781) (#57789) 

Improve efficiency of background indexer by allowing to add
an assertion for failures while they are produced to prevent
queuing them up.
Also, add non-blocking stop to the background indexer so that when
stopping multiple indexers we don't needlessly continue indexing
on some indexers while stopping another one.

Closes #57766
This commit is contained in:
Armin Braun 2020-06-08 10:18:47 +02:00 committed by GitHub
parent 6c93fed204
commit 619e4f8c02
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 69 additions and 46 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
server/src/internalClusterTest/java/org/elasticsearch/discovery/DiskDisruptionIT.java
View File

@ -137,7 +137,7 @@ public class DiskDisruptionIT extends AbstractDisruptionTestCase {
false, random())) {
indexer.setRequestTimeout(TimeValue.ZERO);
indexer.setIgnoreIndexingFailures(true);
indexer.setAssertNoFailuresOnStop(false);
indexer.setFailureAssertion(e -> {});
indexer.start(-1);
waitForDocs(randomIntBetween(1, 100), indexer);

15
server/src/internalClusterTest/java/org/elasticsearch/indices/state/CloseIndexIT.java
View File

@ -201,19 +201,12 @@ public class CloseIndexIT extends ESIntegTestCase {
int nbDocs = 0;
try (BackgroundIndexer indexer = new BackgroundIndexer(indexName, "_doc", client(), MAX_DOCS)) {
indexer.setAssertNoFailuresOnStop(false);
indexer.setFailureAssertion(t -> assertException(t, indexName));
waitForDocs(randomIntBetween(10, 50), indexer);
assertBusy(() -> closeIndices(indexName));
indexer.stop();
indexer.stopAndAwaitStopped();
nbDocs += indexer.totalIndexedDocs();
final Throwable[] failures = indexer.getFailures();
if (failures != null) {
for (Throwable failure : failures) {
assertException(failure, indexName);
}
}
}
assertIndexIsClosed(indexName);
@ -280,6 +273,7 @@ public class CloseIndexIT extends ESIntegTestCase {
createIndex(indexName);
final BackgroundIndexer indexer = new BackgroundIndexer(indexName, "_doc", client(), MAX_DOCS);
indexer.setFailureAssertion(e -> {});
waitForDocs(1, indexer);
final CountDownLatch latch = new CountDownLatch(1);
@ -321,8 +315,7 @@ public class CloseIndexIT extends ESIntegTestCase {
thread.join();
}
indexer.setAssertNoFailuresOnStop(false);
indexer.stop();
indexer.stopAndAwaitStopped();
final ClusterState clusterState = client().admin().cluster().prepareState().get().getState();
if (clusterState.metadata().indices().get(indexName).getState() == IndexMetadata.State.CLOSE) {

16
server/src/internalClusterTest/java/org/elasticsearch/indices/state/CloseWhileRelocatingShardsIT.java
View File

@ -112,6 +112,7 @@ public class CloseWhileRelocatingShardsIT extends ESIntegTestCase {
logger.debug("creating index {} with background indexing", indexName);
final BackgroundIndexer indexer = new BackgroundIndexer(indexName, "_doc", client(), -1, 1);
indexers.put(indexName, indexer);
indexer.setFailureAssertion(t -> assertException(t, indexName));
waitForDocs(1, indexer);
}
docsPerIndex.put(indexName, (long) nbDocs);
@ -225,20 +226,15 @@ public class CloseWhileRelocatingShardsIT extends ESIntegTestCase {
thread.join();
}
// stop indexers first without waiting for stop to not redundantly index on some while waiting for another one to stop
for (BackgroundIndexer indexer : indexers.values()) {
indexer.stop();
}
for (Map.Entry<String, BackgroundIndexer> entry : indexers.entrySet()) {
final BackgroundIndexer indexer = entry.getValue();
indexer.setAssertNoFailuresOnStop(false);
indexer.stop();
indexer.awaitStopped();
final String indexName = entry.getKey();
docsPerIndex.computeIfPresent(indexName, (key, value) -> value + indexer.totalIndexedDocs());
final Throwable[] failures = indexer.getFailures();
if (failures != null) {
for (Throwable failure : failures) {
assertException(failure, indexName);
}
}
}
for (String index : indices) {

8
server/src/internalClusterTest/java/org/elasticsearch/recovery/RecoveryWhileUnderLoadIT.java
View File

@ -128,7 +128,7 @@ public class RecoveryWhileUnderLoadIT extends ESIntegTestCase {
logger.info("--> {} docs indexed", totalNumDocs);
logger.info("--> marking and waiting for indexing threads to stop ...");
indexer.stop();
indexer.stopAndAwaitStopped();
logger.info("--> indexing threads stopped");
logger.info("--> refreshing the index");
@ -183,7 +183,7 @@ public class RecoveryWhileUnderLoadIT extends ESIntegTestCase {
logger.info("--> {} docs indexed", totalNumDocs);
logger.info("--> marking and waiting for indexing threads to stop ...");
indexer.stop();
indexer.stopAndAwaitStopped();
logger.info("--> indexing threads stopped");
logger.info("--> refreshing the index");
@ -261,7 +261,7 @@ public class RecoveryWhileUnderLoadIT extends ESIntegTestCase {
.setWaitForNoRelocatingShards(true));
logger.info("--> marking and waiting for indexing threads to stop ...");
indexer.stop();
indexer.stopAndAwaitStopped();
logger.info("--> indexing threads stopped");
assertNoTimeout(client().admin().cluster().prepareHealth()
@ -302,7 +302,7 @@ public class RecoveryWhileUnderLoadIT extends ESIntegTestCase {
}
logger.info("--> marking and waiting for indexing threads to stop ...");
indexer.stop();
indexer.stopAndAwaitStopped();
logger.info("--> indexing threads stopped");
logger.info("--> bump up number of replicas to 1 and allow all nodes to hold the index");

2
server/src/internalClusterTest/java/org/elasticsearch/recovery/RelocationIT.java
View File

@ -236,7 +236,7 @@ public class RelocationIT extends ESIntegTestCase {
}
logger.info("--> done relocations");
logger.info("--> waiting for indexing threads to stop ...");
indexer.stop();
indexer.stopAndAwaitStopped();
logger.info("--> indexing threads stopped");
logger.info("--> refreshing the index");

70
test/framework/src/main/java/org/elasticsearch/test/BackgroundIndexer.java
View File

@ -39,14 +39,16 @@ import org.elasticsearch.common.xcontent.XContentFactory;
import org.junit.Assert;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Random;
import java.util.Set;
import java.util.concurrent.CopyOnWriteArrayList;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.Semaphore;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicLong;
import java.util.function.Consumer;
import static org.hamcrest.Matchers.emptyIterable;
import static org.hamcrest.Matchers.equalTo;
@ -58,7 +60,7 @@ public class BackgroundIndexer implements AutoCloseable {
final Thread[] writers;
final Client client;
final CountDownLatch stopLatch;
final CopyOnWriteArrayList<Exception> failures;
final Collection<Exception> failures = new ArrayList<>();
final AtomicBoolean stop = new AtomicBoolean(false);
final AtomicLong idGenerator = new AtomicLong();
final CountDownLatch startLatch = new CountDownLatch(1);
@ -66,7 +68,7 @@ public class BackgroundIndexer implements AutoCloseable {
final Semaphore availableBudget = new Semaphore(0);
final boolean useAutoGeneratedIDs;
private final Set<String> ids = ConcurrentCollections.newConcurrentSet();
private boolean assertNoFailuresOnStop = true;
private volatile Consumer<Exception> failureAssertion = null;
volatile int minFieldSize = 10;
volatile int maxFieldSize = 140;
@ -118,7 +120,6 @@ public class BackgroundIndexer implements AutoCloseable {
}
this.client = client;
useAutoGeneratedIDs = random.nextBoolean();
failures = new CopyOnWriteArrayList<>();
writers = new Thread[writerCount];
stopLatch = new CountDownLatch(writers.length);
logger.info("--> creating {} indexing threads (auto start: [{}], numOfDocs: [{}])", writerCount, autoStart, numOfDocs);
@ -162,7 +163,7 @@ public class BackgroundIndexer implements AutoCloseable {
boolean add = ids.add(bulkItemResponse.getId());
assert add : "ID: " + bulkItemResponse.getId() + " already used";
} else {
failures.add(bulkItemResponse.getFailure().getCause());
trackFailure(bulkItemResponse.getFailure().getCause());
}
}
} catch (Exception e) {
@ -204,7 +205,7 @@ public class BackgroundIndexer implements AutoCloseable {
}
logger.info("**** done indexing thread {} stop: {} numDocsIndexed: {}", indexerId, stop.get(), ids.size());
} catch (Exception e) {
failures.add(e);
trackFailure(e);
final long docId = id;
logger.warn(
(Supplier<?>)
@ -222,6 +223,16 @@ public class BackgroundIndexer implements AutoCloseable {
}
}
private void trackFailure(Exception e) {
synchronized (failures) {
if (failureAssertion != null) {
failureAssertion.accept(e);
} else {
failures.add(e);
}
}
}
private XContentBuilder generateSource(long id, Random random) throws IOException {
int contentLength = RandomNumbers.randomIntBetween(random, minFieldSize, maxFieldSize);
StringBuilder text = new StringBuilder(contentLength);
@ -287,32 +298,55 @@ public class BackgroundIndexer implements AutoCloseable {
setBudget(numOfDocs);
}
/** Stop all background threads * */
public void stop() throws InterruptedException {
if (stop.get()) {
return;
}
/** Stop all background threads but don't wait for ongoing indexing operations to finish * */
public void stop() {
stop.set(true);
}
public void awaitStopped() throws InterruptedException {
assert stop.get();
Assert.assertThat("timeout while waiting for indexing threads to stop", stopLatch.await(6, TimeUnit.MINUTES), equalTo(true));
if (assertNoFailuresOnStop) {
if (failureAssertion == null) {
assertNoFailures();
}
}
/** Stop all background threads and wait for ongoing indexing operations to finish * */
public void stopAndAwaitStopped() throws InterruptedException {
stop();
awaitStopped();
}
public long totalIndexedDocs() {
return ids.size();
}
public Throwable[] getFailures() {
return failures.toArray(new Throwable[failures.size()]);
}
public void assertNoFailures() {
synchronized (failures) {
Assert.assertThat(failures, emptyIterable());
}
}
public void setAssertNoFailuresOnStop(final boolean assertNoFailuresOnStop) {
this.assertNoFailuresOnStop = assertNoFailuresOnStop;
/**
* Set a consumer that can be used to run assertions on failures during indexing. If such a consumer is set then it disables adding
* failures to {@link #failures}. Should be used if the number of expected failures during indexing could become very large.
*/
public void setFailureAssertion(Consumer<Exception> failureAssertion) {
synchronized (failures) {
this.failureAssertion = failureAssertion;
boolean success = false;
try {
for (Exception failure : failures) {
failureAssertion.accept(failure);
}
failures.clear();
success = true;
} finally {
if (success == false) {
stop();
}
}
}
}
@Override