KAFKA-12738: track processing errors and implement constant-time task backoff #11787
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,98 @@ | ||
| /* | ||
| * 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.kafka.streams.processor.internals; | ||
|
|
||
| import org.apache.kafka.common.utils.LogContext; | ||
| import org.apache.kafka.streams.processor.TaskId; | ||
|
|
||
| import java.util.Map; | ||
| import java.util.Set; | ||
| import java.util.concurrent.ConcurrentHashMap; | ||
| import org.slf4j.Logger; | ||
|
|
||
| import static org.apache.kafka.streams.processor.internals.TopologyMetadata.UNNAMED_TOPOLOGY; | ||
|
|
||
| /** | ||
| * Multi-threaded class that tracks the status of active tasks being processed. A single instance of this class is | ||
| * shared between all StreamThreads. | ||
| */ | ||
| public class TaskExecutionMetadata { | ||
| private final boolean hasNamedTopologies; | ||
| // map of topologies experiencing errors/currently under backoff | ||
| private final ConcurrentHashMap<String, NamedTopologyMetadata> topologyNameToErrorMetadata = new ConcurrentHashMap<>(); | ||
|
|
||
| public TaskExecutionMetadata(final Set<String> allTopologyNames) { | ||
| this.hasNamedTopologies = !(allTopologyNames.size() == 1 && allTopologyNames.contains(UNNAMED_TOPOLOGY)); | ||
| } | ||
|
|
||
| public boolean canProcessTask(final Task task, final long now) { | ||
| final String topologyName = task.id().topologyName(); | ||
| if (!hasNamedTopologies) { | ||
| // TODO implement error handling/backoff for non-named topologies (needs KIP) | ||
| return true; | ||
| } else { | ||
| final NamedTopologyMetadata metadata = topologyNameToErrorMetadata.get(topologyName); | ||
| return metadata == null || (metadata.canProcess() && metadata.canProcessTask(task, now)); | ||
| } | ||
| } | ||
|
|
||
| public void registerTaskError(final Task task, final Throwable t, final long now) { | ||
| if (hasNamedTopologies) { | ||
| final String topologyName = task.id().topologyName(); | ||
| topologyNameToErrorMetadata.computeIfAbsent(topologyName, n -> new NamedTopologyMetadata(topologyName)) | ||
|
Member
Author
There was a problem hiding this comment. Changed it so that we only track topologies in the map here if they have an active backoff/task in error, rather than registering and unregistering named topologies and trying to keep this in sync between the TopologyMetadata and the individual StreamThreads' view (which was starting to look pretty ugly) Instead we just pop the topology's metadata into the map when one of its tasks hits a new error, and clear it if/when all tasks are healthy again |
||
| .registerTaskError(task, t, now); | ||
| } | ||
| } | ||
|
|
||
| class NamedTopologyMetadata { | ||
ableegoldman marked this conversation as resolved.
Show resolved
Hide resolved
|
||
| private final Logger log; | ||
| private final Map<TaskId, Long> tasksToErrorTime = new ConcurrentHashMap<>(); | ||
|
|
||
| public NamedTopologyMetadata(final String topologyName) { | ||
| final LogContext logContext = new LogContext(String.format("topology-name [%s] ", topologyName)); | ||
| this.log = logContext.logger(NamedTopologyMetadata.class); | ||
| } | ||
|
|
||
| public boolean canProcess() { | ||
| // TODO: during long task backoffs, pause the full topology to avoid it getting out of sync | ||
| return true; | ||
| } | ||
|
|
||
| public boolean canProcessTask(final Task task, final long now) { | ||
| // TODO: implement exponential backoff, for now we just wait 15s | ||
| final Long errorTime = tasksToErrorTime.get(task.id()); | ||
| if (errorTime == null) { | ||
| return true; | ||
| } else if (now - errorTime > 15000L) { | ||
|
Contributor
There was a problem hiding this comment. Curious why the magic number of 15s?
Member
Author
There was a problem hiding this comment. Because it was actually taking the thread 10s to come back up (in the integration test where we overrode Now with that fix it takes .5 - 4s for the thread to be replaced, so there's no particular reason to have it be 15s. I think it makes sense to lower it to maybe 5s for now, and then when we have the true exponential backoff obviously it can start lower and grow from there. |
||
| log.info("End backoff for task {} at t={}", task.id(), now); | ||
| tasksToErrorTime.remove(task.id()); | ||
| if (tasksToErrorTime.isEmpty()) { | ||
| topologyNameToErrorMetadata.remove(task.id().topologyName()); | ||
| } | ||
| return true; | ||
| } else { | ||
| log.debug("Skipping processing for unhealthy task {} at t={}", task.id(), now); | ||
| return false; | ||
| } | ||
| } | ||
|
|
||
| public synchronized void registerTaskError(final Task task, final Throwable t, final long now) { | ||
| log.info("Begin backoff for unhealthy task {} at t={}", task.id(), now); | ||
| tasksToErrorTime.put(task.id(), now); | ||
| } | ||
| } | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure whether it is or will be cleaner in the long run to have this separate class that now has to keep up with topology additions/removals vs just doing all this bookkeeping inside the TopologyMetadata/InternalTopologyBuilder classes -- but until we can carve out time for a real tech debt cleanup of those classes which are already pretty out of control, I felt it best to pull everything out even if it meant duplicated un/registration of topologies
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for bringing this up, I think we can come back and clean this up after we've gained confidence and is ready to extend beyond named topology later.