voonhous commented on code in PR #19575:
URL: https://github.com/apache/hudi/pull/19575#discussion_r3863300953
##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/client/common/HoodieSparkEngineContext.java:
##########
@@ -278,11 +279,43 @@ public String getApplicationId() {
return javaSparkContext.sc().applicationId();
}
+ /**
+ * Drops a registry from both process-wide maps. Only for tests that create
their own SparkContexts:
+ * without it they leave accumulators bound to stopped contexts behind for
whatever runs next in the
+ * same JVM.
+ */
+ @VisibleForTesting
+ public static void removeMetricRegistryForTesting(String tableName, String
registryName) {
+ DISTRIBUTED_REGISTRY_MAP.remove(tableName.isEmpty() ? registryName :
tableName + "." + registryName);
+ Registry.REGISTRY_MAP.remove(Registry.makeKey(tableName, registryName));
+ }
+
@Override
public Registry getMetricRegistry(String tableName, String registryName) {
Review Comment:
Filed as #19759 with the deferred items listed; resolving on that basis.
##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/index/RecordIndexLookupMetrics.java:
##########
@@ -0,0 +1,117 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.hudi.index;
+
+import org.apache.hudi.common.engine.HoodieEngineContext;
+import org.apache.hudi.common.metrics.Registry;
+import org.apache.hudi.config.HoodieWriteConfig;
+import org.apache.hudi.metrics.DistributedRegistry;
+import org.apache.hudi.metrics.ExecutorMetricGroup;
+import org.apache.hudi.metrics.ExecutorMetricRegistry;
+import org.apache.hudi.metrics.RecordIndexMetricNames;
+
+import java.util.Arrays;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.Map;
+import java.util.Set;
+
+/** Executor-side emission for the record index lookup counters. */
+public class RecordIndexLookupMetrics {
+
+ /** Set by the read client around its own tagging call, so dedupe traffic is
attributable separately. */
+ private static final ThreadLocal<String> CALLER =
+ ThreadLocal.withInitial(() ->
RecordIndexMetricNames.CALLER_TAG_LOCATION);
+
+ private RecordIndexLookupMetrics() {
+ }
+
+ public static String currentCaller() {
+ return CALLER.get();
+ }
+
+ /** Restore rather than clear, so a nested tagging call does not reset the
label. */
+ public static String setCaller(String caller) {
+ String previous = CALLER.get();
+ CALLER.set(caller);
+ return previous;
+ }
+
+ public static void restoreCaller(String previous) {
+ CALLER.set(previous);
+ }
+
+ /**
+ * The registries a lookup task collects into, keyed by bare name. Includes
every entry on
+ * {@link ExecutorMetricRegistry}. Delivery is by closure capture, which is
deterministic; resolution is
+ * by name, which lets code below the write API take part without a
signature change.
+ */
+ public static Map<String, Registry> resolveBundle(HoodieEngineContext
context, HoodieWriteConfig config) {
+ return resolveBundle(context, config,
Arrays.asList(ExecutorMetricRegistry.values()));
+ }
+
+ /** Visible for testing the bundle against a group the enum does not ship
with. */
+ public static Map<String, Registry> resolveBundle(HoodieEngineContext
context, HoodieWriteConfig config,
+ Collection<? extends
ExecutorMetricGroup> groups) {
+ Map<String, Registry> bundle = new HashMap<>();
+ for (ExecutorMetricGroup metricRegistry : groups) {
+ if (!metricRegistry.isEnabled(config)) {
+ continue;
+ }
+ Registry registry = context.getMetricRegistry(config.getTableName(),
+ metricRegistry.scopedName(config.getBasePath()));
+ // Only the accumulator-backed registry aggregates back to the driver,
so anything else is left out
+ // rather than bound: a bound LocalRegistry would collect on the
executor and be dropped on the floor,
+ // whereas leaving it out makes the lookup resolve to a no-op that
reports nothing.
+ if (registry instanceof DistributedRegistry) {
+ bundle.put(metricRegistry.registryName(), registry);
+ }
+ }
+ return bundle.isEmpty() ? Collections.emptyMap() : bundle;
+ }
+
+ /**
+ * Records one shard's lookup outcome. Counts records rather than distinct
keys, so
+ * {@code hits + misses == records_looked_up} holds when a batch repeats a
key. Membership is tested
+ * against the found set, bounded by the hit count, not the asked-about set,
bounded by shard size.
+ *
+ * @param keysLookedUp every record key routed to this shard
+ * @param foundKeys the subset present in the index
+ * @param elapsedMs wall-clock spent reading this shard
+ */
+ public static void recordShardLookup(String caller, Collection<String>
keysLookedUp,
+ Collection<String> foundKeys, long
elapsedMs) {
+ if (keysLookedUp.isEmpty()) {
+ return;
+ }
+ Registry registry =
Registry.getRegistry(RecordIndexMetricNames.REGISTRY_NAME);
Review Comment:
Filed as #19759 with the deferred items listed; resolving on that basis.
--
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]
