Skip to content

Commit

Permalink
Query-level resource usages tracking (opensearch-project#13172)
Browse files Browse the repository at this point in the history
* Query-level resource usages tracking

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* Moving TaskResourceTrackingService to clusterService

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* use shard response header to piggyback task resource usages

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* split changes for query insights plugin

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* improve the supplier logic and other misc items

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* track resource usage for failed requests

Signed-off-by: Chenyang Ji <cyji@amazon.com>

* move resource usages interactions into TaskResourceTrackingService

Signed-off-by: Chenyang Ji <cyji@amazon.com>

---------

Signed-off-by: Chenyang Ji <cyji@amazon.com>
  • Loading branch information
ansjcy committed Jun 7, 2024
1 parent cfaa196 commit 3d1fa98
Show file tree
Hide file tree
Showing 29 changed files with 763 additions and 66 deletions.
1 change: 1 addition & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
- [Remote Store] Add support to disable flush based on translog reader count ([#14027](https://github.com/opensearch-project/OpenSearch/pull/14027))
- [Query Insights] Add exporter support for top n queries ([#12982](https://github.com/opensearch-project/OpenSearch/pull/12982))
- [Query Insights] Add X-Opaque-Id to search request metadata for top n queries ([#13374](https://github.com/opensearch-project/OpenSearch/pull/13374))
- Add support for query level resource usage tracking ([#13172](https://github.com/opensearch-project/OpenSearch/pull/13172))

### Dependencies
- Bump `com.github.spullara.mustache.java:compiler` from 0.9.10 to 0.9.13 ([#13329](https://github.com/opensearch-project/OpenSearch/pull/13329), [#13559](https://github.com/opensearch-project/OpenSearch/pull/13559))
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,10 @@ public long getTotalValue() {
return endValue.get() - startValue;
}

public long getStartValue() {
return startValue;
}

@Override
public String toString() {
return String.valueOf(getTotalValue());
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,225 @@
/*
* SPDX-License-Identifier: Apache-2.0
*
* The OpenSearch Contributors require contributions made to
* this file be licensed under the Apache-2.0 license or a
* compatible open source license.
*/

package org.opensearch.core.tasks.resourcetracker;

import org.opensearch.common.annotation.PublicApi;
import org.opensearch.core.ParseField;
import org.opensearch.core.common.Strings;
import org.opensearch.core.common.io.stream.StreamInput;
import org.opensearch.core.common.io.stream.StreamOutput;
import org.opensearch.core.common.io.stream.Writeable;
import org.opensearch.core.xcontent.ConstructingObjectParser;
import org.opensearch.core.xcontent.MediaTypeRegistry;
import org.opensearch.core.xcontent.ToXContentObject;
import org.opensearch.core.xcontent.XContentBuilder;

import java.io.IOException;
import java.util.Objects;

import static org.opensearch.core.xcontent.ConstructingObjectParser.constructorArg;

/**
* Task resource usage information with minimal information about the task
* <p>
* Writeable TaskResourceInfo objects are used to represent resource usage
* information of running tasks, which can be propagated to coordinator node
* to infer query-level resource usage
*
* @opensearch.api
*/
@PublicApi(since = "2.15.0")
public class TaskResourceInfo implements Writeable, ToXContentObject {
private final String action;
private final long taskId;
private final long parentTaskId;
private final String nodeId;
private final TaskResourceUsage taskResourceUsage;

private static final ParseField ACTION = new ParseField("action");
private static final ParseField TASK_ID = new ParseField("taskId");
private static final ParseField PARENT_TASK_ID = new ParseField("parentTaskId");
private static final ParseField NODE_ID = new ParseField("nodeId");
private static final ParseField TASK_RESOURCE_USAGE = new ParseField("taskResourceUsage");

public TaskResourceInfo(
final String action,
final long taskId,
final long parentTaskId,
final String nodeId,
final TaskResourceUsage taskResourceUsage
) {
this.action = action;
this.taskId = taskId;
this.parentTaskId = parentTaskId;
this.nodeId = nodeId;
this.taskResourceUsage = taskResourceUsage;
}

public static final ConstructingObjectParser<TaskResourceInfo, Void> PARSER = new ConstructingObjectParser<>(
"task_resource_info",
a -> new Builder().setAction((String) a[0])
.setTaskId((Long) a[1])
.setParentTaskId((Long) a[2])
.setNodeId((String) a[3])
.setTaskResourceUsage((TaskResourceUsage) a[4])
.build()
);

static {
PARSER.declareString(constructorArg(), ACTION);
PARSER.declareLong(constructorArg(), TASK_ID);
PARSER.declareLong(constructorArg(), PARENT_TASK_ID);
PARSER.declareString(constructorArg(), NODE_ID);
PARSER.declareObject(constructorArg(), TaskResourceUsage.PARSER, TASK_RESOURCE_USAGE);
}

@Override
public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
builder.startObject();
builder.field(ACTION.getPreferredName(), this.action);
builder.field(TASK_ID.getPreferredName(), this.taskId);
builder.field(PARENT_TASK_ID.getPreferredName(), this.parentTaskId);
builder.field(NODE_ID.getPreferredName(), this.nodeId);
builder.startObject(TASK_RESOURCE_USAGE.getPreferredName());
this.taskResourceUsage.toXContent(builder, params);
builder.endObject();
builder.endObject();
return builder;
}

/**
* Builder for {@link TaskResourceInfo}
*/
public static class Builder {
private TaskResourceUsage taskResourceUsage;
private String action;
private long taskId;
private long parentTaskId;
private String nodeId;

public Builder setTaskResourceUsage(final TaskResourceUsage taskResourceUsage) {
this.taskResourceUsage = taskResourceUsage;
return this;
}

public Builder setAction(final String action) {
this.action = action;
return this;
}

public Builder setTaskId(final long taskId) {
this.taskId = taskId;
return this;
}

public Builder setParentTaskId(final long parentTaskId) {
this.parentTaskId = parentTaskId;
return this;
}

public Builder setNodeId(final String nodeId) {
this.nodeId = nodeId;
return this;
}

public TaskResourceInfo build() {
return new TaskResourceInfo(action, taskId, parentTaskId, nodeId, taskResourceUsage);
}
}

/**
* Read task info from a stream.
*
* @param in StreamInput to read
* @return {@link TaskResourceInfo}
* @throws IOException IOException
*/
public static TaskResourceInfo readFromStream(StreamInput in) throws IOException {
return new TaskResourceInfo.Builder().setAction(in.readString())
.setTaskId(in.readLong())
.setParentTaskId(in.readLong())
.setNodeId(in.readString())
.setTaskResourceUsage(TaskResourceUsage.readFromStream(in))
.build();
}

/**
* Get TaskResourceUsage
*
* @return taskResourceUsage
*/
public TaskResourceUsage getTaskResourceUsage() {
return taskResourceUsage;
}

/**
* Get parent task id
*
* @return parent task id
*/
public long getParentTaskId() {
return parentTaskId;
}

/**
* Get task id
* @return task id
*/
public long getTaskId() {
return taskId;
}

/**
* Get node id
* @return node id
*/
public String getNodeId() {
return nodeId;
}

/**
* Get task action
* @return task action
*/
public String getAction() {
return action;
}

@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeString(action);
out.writeLong(taskId);
out.writeLong(parentTaskId);
out.writeString(nodeId);
taskResourceUsage.writeTo(out);
}

@Override
public String toString() {
return Strings.toString(MediaTypeRegistry.JSON, this);
}

@Override
public boolean equals(Object obj) {
if (obj == null || obj.getClass() != TaskResourceInfo.class) {
return false;
}
TaskResourceInfo other = (TaskResourceInfo) obj;
return action.equals(other.action)
&& taskId == other.taskId
&& parentTaskId == other.parentTaskId
&& Objects.equals(nodeId, other.nodeId)
&& taskResourceUsage.equals(other.taskResourceUsage);
}

@Override
public int hashCode() {
return Objects.hash(action, taskId, parentTaskId, nodeId, taskResourceUsage);
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@
import org.opensearch.core.action.ActionListener;
import org.opensearch.core.action.ShardOperationFailedException;
import org.opensearch.core.index.shard.ShardId;
import org.opensearch.core.tasks.resourcetracker.TaskResourceInfo;
import org.opensearch.search.SearchPhaseResult;
import org.opensearch.search.SearchShardTarget;
import org.opensearch.search.internal.AliasFilter;
Expand Down Expand Up @@ -469,6 +470,10 @@ private void onRequestEnd(SearchRequestContext searchRequestContext) {
this.searchRequestContext.getSearchRequestOperationsListener().onRequestEnd(this, searchRequestContext);
}

private void onRequestFailure(SearchRequestContext searchRequestContext) {
this.searchRequestContext.getSearchRequestOperationsListener().onRequestFailure(this, searchRequestContext);
}

private void executePhase(SearchPhase phase) {
Span phaseSpan = tracer.startSpan(SpanCreationContext.server().name("[phase/" + phase.getName() + "]"));
try (final SpanScope scope = tracer.withSpanInScope(phaseSpan)) {
Expand Down Expand Up @@ -507,6 +512,7 @@ ShardSearchFailure[] buildShardFailures() {
private void onShardFailure(final int shardIndex, @Nullable SearchShardTarget shard, final SearchShardIterator shardIt, Exception e) {
// we always add the shard failure for a specific shard instance
// we do make sure to clean it on a successful response from a shard
setPhaseResourceUsages();
onShardFailure(shardIndex, shard, e);
SearchShardTarget nextShard = FailAwareWeightedRouting.getInstance()
.findNext(shardIt, clusterState, e, () -> totalOps.incrementAndGet());
Expand Down Expand Up @@ -618,9 +624,15 @@ protected void onShardResult(Result result, SearchShardIterator shardIt) {
if (logger.isTraceEnabled()) {
logger.trace("got first-phase result from {}", result != null ? result.getSearchShardTarget() : null);
}
this.setPhaseResourceUsages();
results.consumeResult(result, () -> onShardResultConsumed(result, shardIt));
}

public void setPhaseResourceUsages() {
TaskResourceInfo taskResourceUsage = searchRequestContext.getTaskResourceUsageSupplier().get();
searchRequestContext.recordPhaseResourceUsage(taskResourceUsage);
}

private void onShardResultConsumed(Result result, SearchShardIterator shardIt) {
successfulOps.incrementAndGet();
// clean a previous error on this shard group (note, this code will be serialized on the same shardIndex value level
Expand Down Expand Up @@ -751,6 +763,7 @@ public void sendSearchResponse(InternalSearchResponse internalSearchResponse, At

@Override
public final void onPhaseFailure(SearchPhase phase, String msg, Throwable cause) {
setPhaseResourceUsages();
if (currentPhaseHasLifecycle) {
this.searchRequestContext.getSearchRequestOperationsListener().onPhaseFailure(this, cause);
}
Expand Down Expand Up @@ -780,6 +793,7 @@ private void raisePhaseFailure(SearchPhaseExecutionException exception) {
});
}
Releasables.close(releasables);
onRequestFailure(searchRequestContext);
listener.onFailure(exception);
}

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -240,6 +240,7 @@ private void executeFetch(
public void innerOnResponse(FetchSearchResult result) {
try {
progressListener.notifyFetchResult(shardIndex);
context.setPhaseResourceUsages();
counter.onResult(result);
} catch (Exception e) {
context.onPhaseFailure(FetchSearchPhase.this, "", e);
Expand All @@ -254,6 +255,7 @@ public void onFailure(Exception e) {
e
);
progressListener.notifyFetchFailure(shardIndex, shardTarget, e);
context.setPhaseResourceUsages();
counter.onFailure(shardIndex, shardTarget, e);
} finally {
// the search context might not be cleared on the node where the fetch was executed for example
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -150,4 +150,9 @@ default void sendReleaseSearchContext(
* Registers a {@link Releasable} that will be closed when the search request finishes or fails.
*/
void addReleasable(Releasable releasable);

/**
* Set the resource usage info for this phase
*/
void setPhaseResourceUsages();
}
Loading

0 comments on commit 3d1fa98

Please sign in to comment.