Skip to content

[server] Keep snapshot metrics safe during cleanup #4089

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
platinumhamburg wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[server] Keep snapshot metrics safe during cleanup #4089

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 2 additions & 10 deletions ...rver/src/main/java/org/apache/fluss/server/coordinator/CompletedSnapshotStoreManager.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -122,14 +122,6 @@ private long physicalStorageRemoteKvSize() {
.reduce(0L, Long::sum);
}

private long getNumSnapshots(TableBucket tableBucket) {
return bucketCompletedSnapshotStores.get(tableBucket).getNumSnapshots();
}

private long getAllSnapshotSize(TableBucket tableBucket) {
return bucketCompletedSnapshotStores.get(tableBucket).getPhysicalStorageRemoteKvSize();
}

public CompletedSnapshotStore getOrCreateCompletedSnapshotStore(
TablePath tablePath, TableBucket tableBucket) {
return bucketCompletedSnapshotStores.computeIfAbsent(
Expand All @@ -152,10 +144,10 @@ public CompletedSnapshotStore getOrCreateCompletedSnapshotStore(
if (bucketMetricGroup != null) {
LOG.info("Add bucketMetricGroup for tableBucket {}.", bucket);
bucketMetricGroup.gauge(
MetricNames.KV_NUM_SNAPSHOTS, () -> getNumSnapshots(bucket));
MetricNames.KV_NUM_SNAPSHOTS, snapshotStore::getNumSnapshots);
bucketMetricGroup.gauge(
MetricNames.KV_ALL_SNAPSHOT_SIZE,
() -> getAllSnapshotSize(bucket));
snapshotStore::getPhysicalStorageRemoteKvSize);
} else {
LOG.warn(
"Failed to add bucketMetricGroup for tableBucket {} when creating completed snapshot.",
Expand Down
18 changes: 10 additions & 8 deletions ...s-server/src/main/java/org/apache/fluss/server/coordinator/CoordinatorEventProcessor.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1068,9 +1068,13 @@ private void processCreatePartition(CreatePartitionEvent createPartitionEvent) {
}

private void processDropTable(DropTableEvent dropTableEvent) {
// If this is a primary key table, drop the kv snapshot store.
long tableId = dropTableEvent.getTableId();
TableInfo dropTableInfo = coordinatorContext.getTableInfoById(tableId);

// Remove table metrics before dropping their backing snapshot stores.
coordinatorMetricGroup.removeTableMetricGroup(dropTableInfo.getTablePath(), tableId);

// If this is a primary key table, drop the kv snapshot store.
if (dropTableInfo.hasPrimaryKey()) {
Set<TableBucket> deleteTableBuckets = coordinatorContext.getAllBucketsForTable(tableId);
completedSnapshotStoreManager.removeCompletedSnapshotStoreByTableBuckets(
Expand All @@ -1093,9 +1097,6 @@ private void processDropTable(DropTableEvent dropTableEvent) {
null,
Collections.emptySet());

// remove table metrics.
coordinatorMetricGroup.removeTableMetricGroup(dropTableInfo.getTablePath(), tableId);

// For partitioned tables, the dropped table has no table-level replicas
// (all buckets live under partitionAssignments), so getAllReplicasForTable
// returns empty and areAllReplicasInState(.., ReplicaDeletionSuccessful)
Expand All @@ -1115,6 +1116,11 @@ private void processDropPartition(DropPartitionEvent dropPartitionEvent) {

// If this is a primary key table partition, drop the kv snapshot store.
TableInfo dropTableInfo = coordinatorContext.getTableInfoById(tableId);

// Remove partition metrics before dropping their backing snapshot stores.
coordinatorMetricGroup.removeTablePartitionMetricsGroup(
dropTableInfo.getTablePath(), tableId, tablePartition.getPartitionId());

if (dropTableInfo.hasPrimaryKey()) {
Set<TableBucket> deleteTableBuckets =
coordinatorContext.getAllBucketsForPartition(
Expand All @@ -1133,10 +1139,6 @@ private void processDropPartition(DropPartitionEvent dropPartitionEvent) {
tableId,
tablePartition.getPartitionId(),
Collections.emptySet());

// remove partition metrics.
coordinatorMetricGroup.removeTablePartitionMetricsGroup(
dropTableInfo.getTablePath(), tableId, tablePartition.getPartitionId());
}

private void processDeleteReplicaResponseReceived(
Expand Down
107 changes: 107 additions & 0 deletions ...rver/src/test/java/org/apache/fluss/server/coordinator/CoordinatorEventProcessorTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,9 @@
import org.apache.fluss.metadata.TableInfo;
import org.apache.fluss.metadata.TablePartition;
import org.apache.fluss.metadata.TablePath;
import org.apache.fluss.metrics.Gauge;
import org.apache.fluss.metrics.MetricNames;
import org.apache.fluss.metrics.groups.AbstractMetricGroup;
import org.apache.fluss.rpc.gateway.TabletServerGateway;
import org.apache.fluss.rpc.messages.AdjustIsrResponse;
import org.apache.fluss.rpc.messages.ApiMessage;
Expand All @@ -46,6 +49,7 @@
import org.apache.fluss.rpc.messages.NotifyLeaderAndIsrResponse;
import org.apache.fluss.rpc.messages.NotifyRemoteLogOffsetsRequest;
import org.apache.fluss.rpc.messages.UpdateMetadataRequest;
import org.apache.fluss.rpc.messages.UpdateMetadataResponse;
import org.apache.fluss.rpc.protocol.ApiError;
import org.apache.fluss.rpc.protocol.ApiKeys;
import org.apache.fluss.rpc.protocol.Errors;
Expand Down Expand Up @@ -106,6 +110,7 @@

import java.nio.file.Path;
import java.time.Duration;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.HashMap;
Expand All @@ -115,9 +120,11 @@
import java.util.Set;
import java.util.concurrent.CompletableFuture;
import java.util.concurrent.ConcurrentLinkedDeque;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.function.BiConsumer;
import java.util.function.Consumer;
import java.util.function.Function;
import java.util.stream.Collectors;
Expand All @@ -138,6 +145,7 @@
import static org.apache.fluss.testutils.common.CommonTestUtils.retry;
import static org.apache.fluss.testutils.common.CommonTestUtils.waitValue;
import static org.assertj.core.api.Assertions.assertThat;
import static org.assertj.core.api.Assertions.assertThatCode;
import static org.assertj.core.api.Assertions.assertThatThrownBy;

/** Test for {@link CoordinatorEventProcessor}. */
Expand Down Expand Up @@ -785,6 +793,78 @@ void testAddBucketCompletedSnapshot(@TempDir Path tempDir) throws Exception {
.isInstanceOf(InvalidCoordinatorException.class);
}

@Test
void testMetricsRemainCollectableAfterDropPartitionFailure() throws Exception {
eventProcessor.shutdown();

FailingUpdateMetadataChannelManager failingChannelManager =
new FailingUpdateMetadataChannelManager();
testCoordinatorChannelManager = failingChannelManager;
eventProcessor = buildCoordinatorEventProcessor();
eventProcessor.startup();
completedSnapshotStoreManager = eventProcessor.completedSnapshotStoreManager();

TablePath tablePath = TablePath.of(defaultDatabase, "test_metrics_during_drop_partition");
initCoordinatorChannel();
long tableId =
metadataManager.createTable(
tablePath, remoteDataDir, getPartitionedTable(), null, false);
Map<Integer, BucketAssignment> assignments =
generateAssignment(
N_BUCKETS,
REPLICATION_FACTOR,
new TabletServerInfo[] {
new TabletServerInfo(0, "rack0"),
new TabletServerInfo(1, "rack1"),
new TabletServerInfo(2, "rack2")
})
.getBucketAssignments();
PartitionAssignment partitionAssignment = new PartitionAssignment(tableId, assignments);
PartitionIdName partition =
preparePartitionAssignment(tablePath, tableId, partitionAssignment).f0;
verifyPartitionCreated(
new TablePartition(tableId, partition.partitionId),
partitionAssignment,
N_BUCKETS,
REPLICATION_FACTOR);

List<TableBucket> tableBuckets = allTableBuckets(tableId, partition.partitionId, N_BUCKETS);
List<AbstractMetricGroup> bucketMetricGroups = new ArrayList<>();
List<Gauge<?>> inFlightGauges = new ArrayList<>();
for (TableBucket tableBucket : tableBuckets) {
completedSnapshotStoreManager.getOrCreateCompletedSnapshotStore(tablePath, tableBucket);
AbstractMetricGroup bucketMetricGroup =
(AbstractMetricGroup)
TestingMetricGroups.COORDINATOR_METRICS.getTableBucketMetricGroup(
tablePath, tableBucket);
assertThat(bucketMetricGroup).isNotNull();
bucketMetricGroups.add(bucketMetricGroup);
inFlightGauges.add(
(Gauge<?>) bucketMetricGroup.getMetrics().get(MetricNames.KV_NUM_SNAPSHOTS));
inFlightGauges.add(
(Gauge<?>)
bucketMetricGroup.getMetrics().get(MetricNames.KV_ALL_SNAPSHOT_SIZE));
}
assertThat(inFlightGauges).doesNotContainNull();

failingChannelManager.failNextUpdateMetadata();
zookeeperClient.deletePartition(tablePath, partition.partitionName);
failingChannelManager.awaitFailure();
fromCtx(context -> null);

assertThat(completedSnapshotStoreManager.getBucketCompletedSnapshotStores()).isEmpty();
assertThat(bucketMetricGroups).allMatch(AbstractMetricGroup::isClosed);
assertThat(tableBuckets)
.allSatisfy(
tableBucket ->
assertThat(
TestingMetricGroups.COORDINATOR_METRICS
.getTableBucketMetricGroup(
tablePath, tableBucket))
.isNull());
assertThatCode(() -> inFlightGauges.forEach(Gauge::getValue)).doesNotThrowAnyException();
}

@Test
void testCreateAndDropPartition() throws Exception {
TablePath tablePath = TablePath.of(defaultDatabase, "test_create_drop_partition");
Expand Down Expand Up @@ -2192,6 +2272,33 @@ private CoordinatorEventProcessor buildCoordinatorEventProcessor() {
SystemClock.getInstance());
}

private static class FailingUpdateMetadataChannelManager extends TestCoordinatorChannelManager {

private final CountDownLatch failureObserved = new CountDownLatch(1);
private volatile boolean failUpdateMetadata;

private void failNextUpdateMetadata() {
failUpdateMetadata = true;
}

@Override
public void sendUpdateMetadataRequest(
int serverId,
UpdateMetadataRequest request,
BiConsumer<UpdateMetadataResponse, ? super Throwable> responseConsumer) {
if (failUpdateMetadata) {
failUpdateMetadata = false;
failureObserved.countDown();
throw new RuntimeException("Injected update metadata failure");
}
super.sendUpdateMetadataRequest(serverId, request, responseConsumer);
}

private void awaitFailure() throws InterruptedException {
assertThat(failureObserved.await(30, TimeUnit.SECONDS)).isTrue();
}
}

private static class RecordingAutoPartitionManager extends AutoPartitionManager {

private final CoordinatorMetadataCache metadataCache;
Expand Down
Loading