-
Notifications
You must be signed in to change notification settings - Fork 28.2k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[SPARK-24355] Spark external shuffle server improvement to better han…
…dle block fetch requests. ## What changes were proposed in this pull request? Description: Right now, the default server side netty handler threads is 2 * # cores, and can be further configured with parameter spark.shuffle.io.serverThreads. In order to process a client request, it would require one available server netty handler thread. However, when the server netty handler threads start to process ChunkFetchRequests, they will be blocked on disk I/O, mostly due to disk contentions from the random read operations initiated by all the ChunkFetchRequests received from clients. As a result, when the shuffle server is serving many concurrent ChunkFetchRequests, the server side netty handler threads could all be blocked on reading shuffle files, thus leaving no handler thread available to process other types of requests which should all be very quick to process. This issue could potentially be fixed by limiting the number of netty handler threads that could get blocked when processing ChunkFetchRequest. We have a patch to do this by using a separate EventLoopGroup with a dedicated ChannelHandler to process ChunkFetchRequest. This enables shuffle server to reserve netty handler threads for non-ChunkFetchRequest, thus enabling consistent processing time for these requests which are fast to process. After deploying the patch in our infrastructure, we no longer see timeout issues with either executor registration with local shuffle server or shuffle client establishing connection with remote shuffle server. (Please fill in changes proposed in this fix) For Original PR please refer here #21402 ## How was this patch tested? Unit tests and stress testing. (Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests) (If this patch involves UI changes, please attach a screenshot; otherwise, remove this) Please review http://spark.apache.org/contributing.html before opening a pull request. Closes #22173 from redsanket/SPARK-24335. Authored-by: Sanket Chintapalli <schintap@yahoo-inc.com> Signed-off-by: Thomas Graves <tgraves@apache.org>
- Loading branch information
Showing
9 changed files
with
425 additions
and
88 deletions.
There are no files selected for viewing
This file contains 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
135 changes: 135 additions & 0 deletions
135
...etwork-common/src/main/java/org/apache/spark/network/server/ChunkFetchRequestHandler.java
This file contains 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 +1,135 @@ | ||
/* | ||
* 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.spark.network.server; | ||
|
||
import java.net.SocketAddress; | ||
|
||
import com.google.common.base.Throwables; | ||
import io.netty.channel.Channel; | ||
import io.netty.channel.ChannelFuture; | ||
import io.netty.channel.ChannelFutureListener; | ||
import io.netty.channel.ChannelHandlerContext; | ||
import io.netty.channel.SimpleChannelInboundHandler; | ||
import org.slf4j.Logger; | ||
import org.slf4j.LoggerFactory; | ||
|
||
import org.apache.spark.network.buffer.ManagedBuffer; | ||
import org.apache.spark.network.client.TransportClient; | ||
import org.apache.spark.network.protocol.ChunkFetchFailure; | ||
import org.apache.spark.network.protocol.ChunkFetchRequest; | ||
import org.apache.spark.network.protocol.ChunkFetchSuccess; | ||
import org.apache.spark.network.protocol.Encodable; | ||
|
||
import static org.apache.spark.network.util.NettyUtils.*; | ||
|
||
/** | ||
* A dedicated ChannelHandler for processing ChunkFetchRequest messages. When sending response | ||
* of ChunkFetchRequest messages to the clients, the thread performing the I/O on the underlying | ||
* channel could potentially be blocked due to disk contentions. If several hundreds of clients | ||
* send ChunkFetchRequest to the server at the same time, it could potentially occupying all | ||
* threads from TransportServer's default EventLoopGroup for waiting for disk reads before it | ||
* can send the block data back to the client as part of the ChunkFetchSuccess messages. As a | ||
* result, it would leave no threads left to process other RPC messages, which takes much less | ||
* time to process, and could lead to client timing out on either performing SASL authentication, | ||
* registering executors, or waiting for response for an OpenBlocks messages. | ||
*/ | ||
public class ChunkFetchRequestHandler extends SimpleChannelInboundHandler<ChunkFetchRequest> { | ||
private static final Logger logger = LoggerFactory.getLogger(ChunkFetchRequestHandler.class); | ||
|
||
private final TransportClient client; | ||
private final StreamManager streamManager; | ||
/** The max number of chunks being transferred and not finished yet. */ | ||
private final long maxChunksBeingTransferred; | ||
|
||
public ChunkFetchRequestHandler( | ||
TransportClient client, | ||
StreamManager streamManager, | ||
Long maxChunksBeingTransferred) { | ||
this.client = client; | ||
this.streamManager = streamManager; | ||
this.maxChunksBeingTransferred = maxChunksBeingTransferred; | ||
} | ||
|
||
@Override | ||
public void exceptionCaught(ChannelHandlerContext ctx, Throwable cause) throws Exception { | ||
logger.warn("Exception in connection from " + getRemoteAddress(ctx.channel()), cause); | ||
ctx.close(); | ||
} | ||
|
||
@Override | ||
protected void channelRead0( | ||
ChannelHandlerContext ctx, | ||
final ChunkFetchRequest msg) throws Exception { | ||
Channel channel = ctx.channel(); | ||
if (logger.isTraceEnabled()) { | ||
logger.trace("Received req from {} to fetch block {}", getRemoteAddress(channel), | ||
msg.streamChunkId); | ||
} | ||
long chunksBeingTransferred = streamManager.chunksBeingTransferred(); | ||
if (chunksBeingTransferred >= maxChunksBeingTransferred) { | ||
logger.warn("The number of chunks being transferred {} is above {}, close the connection.", | ||
chunksBeingTransferred, maxChunksBeingTransferred); | ||
channel.close(); | ||
return; | ||
} | ||
ManagedBuffer buf; | ||
try { | ||
streamManager.checkAuthorization(client, msg.streamChunkId.streamId); | ||
streamManager.registerChannel(channel, msg.streamChunkId.streamId); | ||
buf = streamManager.getChunk(msg.streamChunkId.streamId, msg.streamChunkId.chunkIndex); | ||
} catch (Exception e) { | ||
logger.error(String.format("Error opening block %s for request from %s", | ||
msg.streamChunkId, getRemoteAddress(channel)), e); | ||
respond(channel, new ChunkFetchFailure(msg.streamChunkId, | ||
Throwables.getStackTraceAsString(e))); | ||
return; | ||
} | ||
|
||
streamManager.chunkBeingSent(msg.streamChunkId.streamId); | ||
respond(channel, new ChunkFetchSuccess(msg.streamChunkId, buf)).addListener( | ||
(ChannelFutureListener) future -> streamManager.chunkSent(msg.streamChunkId.streamId)); | ||
} | ||
|
||
/** | ||
* The invocation to channel.writeAndFlush is async, and the actual I/O on the | ||
* channel will be handled by the EventLoop the channel is registered to. So even | ||
* though we are processing the ChunkFetchRequest in a separate thread pool, the actual I/O, | ||
* which is the potentially blocking call that could deplete server handler threads, is still | ||
* being processed by TransportServer's default EventLoopGroup. In order to throttle the max | ||
* number of threads that channel I/O for sending response to ChunkFetchRequest, the thread | ||
* calling channel.writeAndFlush will wait for the completion of sending response back to | ||
* client by invoking await(). This will throttle the rate at which threads from | ||
* ChunkFetchRequest dedicated EventLoopGroup submit channel I/O requests to TransportServer's | ||
* default EventLoopGroup, thus making sure that we can reserve some threads in | ||
* TransportServer's default EventLoopGroup for handling other RPC messages. | ||
*/ | ||
private ChannelFuture respond( | ||
final Channel channel, | ||
final Encodable result) throws InterruptedException { | ||
final SocketAddress remoteAddress = channel.remoteAddress(); | ||
return channel.writeAndFlush(result).await().addListener((ChannelFutureListener) future -> { | ||
if (future.isSuccess()) { | ||
logger.trace("Sent result {} to client {}", result, remoteAddress); | ||
} else { | ||
logger.error(String.format("Error sending result %s to %s; closing connection", | ||
result, remoteAddress), future.cause()); | ||
channel.close(); | ||
} | ||
}); | ||
} | ||
} |
This file contains 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 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.