服务器之家:专注于服务器技术及软件下载分享
分类导航

PHP教程|ASP.NET教程|Java教程|ASP教程|编程技术|正则表达式|C/C++|IOS|C#|Swift|Android|VB|R语言|JavaScript|易语言|vb.net|

服务器之家 - 编程语言 - 编程技术 - spark通过kafka-appender指定日志输出到kafka引发的死锁问题

spark通过kafka-appender指定日志输出到kafka引发的死锁问题

2020-09-12 17:22民工哥 编程技术

这篇文章主要介绍了spark通过kafka-appender指定日志输出到kafka引发的死锁,本文通过实例代码给大家介绍的非常详细,对大家的学习或工作具有一定的参考借鉴价值,需要的朋友可以参考下

在采用log4jkafka-appender收集spark任务运行日志时,发现提交到yarn上的任务始终ACCEPTED状态,无法进入RUNNING状态,并且会重试两次后超时。期初认为是yarn资源不足导致,但在确认yarn资源充裕的时候问题依旧,而且基本上能稳定复现。

起初是这么配置spark日志输出到kafka的:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
log4j.rootCategory=INFO, console, kafka
log4j.appender.console=org.apache.log4j.ConsoleAppender
log4j.appender.console.target=System.err
log4j.appender.console.layout=org.apache.log4j.PatternLayout
log4j.appender.console.layout.ConversionPattern=%d{yyyy/MM/dd HH:mm:ss.SSS} %p %c{1}: [${log4j.pipelineId}] %m%n
 
# Kafka appender
log4j.appender.kafka=org.apache.kafka.log4jappender.KafkaLog4jAppender
# Set Kafka topic and brokerList
log4j.appender.kafka.topic=yarn_spark_log
log4j.appender.kafka.brokerList=localhost:9092
log4j.appender.kafka.compressionType=none
log4j.appender.kafka.syncSend=false
log4j.appender.kafka.maxBlockMs=10
log4j.appender.kafka.layout=org.apache.log4j.PatternLayout
log4j.appender.kafka.layout.ConversionPattern=%d{yyyy/MM/dd HH:mm:ss.SSS} %p %c{1}: [${log4j.pipelineId}] %m

这里用org.apache.kafka.log4jappender.KafkaLog4jAppender默认将所有日志都输出到kafka,这个appender已经被kafka官方维护,稳定性应该是可以保障的。

问题定位

发现问题后,尝试将输出到kafka的规则去掉,问题解除!于是把问题定位到跟日志输出到kafka有关。通过其他测试,证实目标kafka其实是正常的,这就非常奇怪了。

查看yarn的ResourceManager日志,发现有如下超时

2020-05-07 21:49:48,230 INFO org.apache.hadoop.yarn.util.AbstractLivelinessMonitor: Expired:appattempt_1578970174552_3204_000002 Timed out after 600 secs
2020-05-07 21:49:48,230 INFO org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl: Updating application attempt appattempt_1578970174552_3204_000002 with final
 state: FAILED, and exit status: -1000
2020-05-07 21:49:48,231 INFO org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl: appattempt_1578970174552_3204_000002 State change from LAUNCHED to FINAL_SAV
ING on event = EXPIRE

表明,yarn本身是接收任务的,但是发现任务迟迟没有启动。在spark的场景下其实是指只有driver启动了,但是没有启动executor。
而查看driver日志,发现日志输出到一个地方就卡住了,不往下继续了。通过对比成功运行和卡住的情况发现,日志卡在这条上:

2020/05/07 19:37:10.324 INFO SecurityManager: Changing view acls to: yarn,root
2020/05/07 19:37:10.344 INFO Metadata: Cluster ID: 6iG6WHA2SoK7FfgGgWHt_A

卡住的情况下,只会打出SecurityManager这行,而无法打出Metadata这行。
猜想Metadata这行是kafka-client本身打出来的,因为整个上下文只有yarn, spark, kafka-client可能会打出这个日志。

在kafka-client 2.2.0版本中找到这个日志是输出位置:

?
1
2
3
4
5
6
7
8
9
public synchronized void update(MetadataResponse metadataResponse, long now) {
  ...
 
  String newClusterId = cache.cluster().clusterResource().clusterId();
  if (!Objects.equals(previousClusterId, newClusterId)) {
    log.info("Cluster ID: {}", newClusterId);
  }
  ...
}

看到synchronized,高度怀疑死锁。于是考虑用jstack分析:

在yarn上运行spark任务的时候,driver进程叫ApplicationMaster,executor进程叫CoarseGrainedExecutorBackend。这里首先尝试再复现过程中找到drvier最终在哪个节点上运行,然后快速使用jstack -F <pid>打印堆栈

jstack果然不负众望,报告了死锁!这里我把结果贴的全一点

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
[root@node1 ~]# jstack 20136
20136: Unable to open socket file: target process not responding or HotSpot VM not loaded
The -F option can be used when the target process is not responding
[root@node1 ~]# jstack -F 20136
Attaching to process ID 20136, please wait...
Debugger attached successfully.
Server compiler detected.
JVM version is 25.231-b11
Deadlock Detection:
 
Found one Java-level deadlock:
=============================
 
"kafka-producer-network-thread | producer-1":
 waiting to lock Monitor@0x00000000025fcc48 (Object@0x00000000ed680b60, a org/apache/kafka/log4jappender/KafkaLog4jAppender),
 which is held by "main"
"main":
 waiting to lock Monitor@0x00007fec9dbde038 (Object@0x00000000ee44de38, a org/apache/kafka/clients/Metadata),
 which is held by "kafka-producer-network-thread | producer-1"
 
Found a total of 1 deadlock.
 
Thread 20157: (state = BLOCKED)
 - org.apache.log4j.AppenderSkeleton.doAppend(org.apache.log4j.spi.LoggingEvent) @bci=0, line=231 (Interpreted frame)
 - org.apache.log4j.helpers.AppenderAttachableImpl.appendLoopOnAppenders(org.apache.log4j.spi.LoggingEvent) @bci=41, line=66 (Interpreted frame)
 - org.apache.log4j.Category.callAppenders(org.apache.log4j.spi.LoggingEvent) @bci=26, line=206 (Interpreted frame)
 - org.apache.log4j.Category.forcedLog(java.lang.String, org.apache.log4j.Priority, java.lang.Object, java.lang.Throwable) @bci=14, line=391 (Interpreted frame)
 - org.apache.log4j.Category.log(java.lang.String, org.apache.log4j.Priority, java.lang.Object, java.lang.Throwable) @bci=34, line=856 (Interpreted frame)
 - org.slf4j.impl.Log4jLoggerAdapter.info(java.lang.String, java.lang.Object) @bci=34, line=324 (Interpreted frame)
 - org.apache.kafka.clients.Metadata.update(org.apache.kafka.common.requests.MetadataResponse, long) @bci=317, line=365 (Interpreted frame)
 - org.apache.kafka.clients.NetworkClient$DefaultMetadataUpdater.handleCompletedMetadataResponse(org.apache.kafka.common.requests.RequestHeader, long, org.apache.kafka.common.requests.MetadataResponse) @bci=184, line=1031 (Interpreted frame)
 - org.apache.kafka.clients.NetworkClient.handleCompletedReceives(java.util.List, long) @bci=215, line=822 (Interpreted frame)
 - org.apache.kafka.clients.NetworkClient.poll(long, long) @bci=132, line=544 (Interpreted frame)
 - org.apache.kafka.clients.producer.internals.Sender.run(long) @bci=227, line=311 (Interpreted frame)
 - org.apache.kafka.clients.producer.internals.Sender.run() @bci=28, line=235 (Interpreted frame)
 - java.lang.Thread.run() @bci=11, line=748 (Interpreted frame)
 
 
Thread 20150: (state = BLOCKED)
 
 
Thread 20149: (state = BLOCKED)
 - java.lang.Object.wait(long) @bci=0 (Interpreted frame)
 - java.lang.ref.ReferenceQueue.remove(long) @bci=59, line=144 (Interpreted frame)
 - java.lang.ref.ReferenceQueue.remove() @bci=2, line=165 (Interpreted frame)
 - java.lang.ref.Finalizer$FinalizerThread.run() @bci=36, line=216 (Interpreted frame)
 
 
Thread 20148: (state = BLOCKED)
 - java.lang.Object.wait(long) @bci=0 (Interpreted frame)
 - java.lang.Object.wait() @bci=2, line=502 (Interpreted frame)
 - java.lang.ref.Reference.tryHandlePending(boolean) @bci=54, line=191 (Interpreted frame)
 - java.lang.ref.Reference$ReferenceHandler.run() @bci=1, line=153 (Interpreted frame)
 
 
Thread 20137: (state = BLOCKED)
 - java.lang.Object.wait(long) @bci=0 (Interpreted frame)
 - org.apache.kafka.clients.Metadata.awaitUpdate(int, long) @bci=63, line=261 (Interpreted frame)
 - org.apache.kafka.clients.producer.KafkaProducer.waitOnMetadata(java.lang.String, java.lang.Integer, long) @bci=160, line=983 (Interpreted frame)
 - org.apache.kafka.clients.producer.KafkaProducer.doSend(org.apache.kafka.clients.producer.ProducerRecord, org.apache.kafka.clients.producer.Callback) @bci=19, line=860 (Interpreted frame)
 - org.apache.kafka.clients.producer.KafkaProducer.send(org.apache.kafka.clients.producer.ProducerRecord, org.apache.kafka.clients.producer.Callback) @bci=12, line=840 (Interpreted frame)
 - org.apache.kafka.clients.producer.KafkaProducer.send(org.apache.kafka.clients.producer.ProducerRecord) @bci=3, line=727 (Interpreted frame)
 - org.apache.kafka.log4jappender.KafkaLog4jAppender.append(org.apache.log4j.spi.LoggingEvent) @bci=69, line=283 (Interpreted frame)
 - org.apache.log4j.AppenderSkeleton.doAppend(org.apache.log4j.spi.LoggingEvent) @bci=106, line=251 (Interpreted frame)
 - org.apache.log4j.helpers.AppenderAttachableImpl.appendLoopOnAppenders(org.apache.log4j.spi.LoggingEvent) @bci=41, line=66 (Interpreted frame)
 - org.apache.log4j.Category.callAppenders(org.apache.log4j.spi.LoggingEvent) @bci=26, line=206 (Interpreted frame)
 - org.apache.log4j.Category.forcedLog(java.lang.String, org.apache.log4j.Priority, java.lang.Object, java.lang.Throwable) @bci=14, line=391 (Interpreted frame)
 - org.apache.log4j.Category.log(java.lang.String, org.apache.log4j.Priority, java.lang.Object, java.lang.Throwable) @bci=34, line=856 (Interpreted frame)
 - org.slf4j.impl.Log4jLoggerAdapter.info(java.lang.String) @bci=12, line=305 (Interpreted frame)
 - org.apache.spark.internal.Logging$class.logInfo(org.apache.spark.internal.Logging, scala.Function0) @bci=29, line=54 (Interpreted frame)
 - org.apache.spark.SecurityManager.logInfo(scala.Function0) @bci=2, line=44 (Interpreted frame)
 - org.apache.spark.SecurityManager.setViewAcls(scala.collection.immutable.Set, java.lang.String) @bci=36, line=139 (Interpreted frame)
 - org.apache.spark.SecurityManager.<init>(org.apache.spark.SparkConf, scala.Option) @bci=158, line=81 (Interpreted frame)
 - org.apache.spark.deploy.yarn.ApplicationMaster.<init>(org.apache.spark.deploy.yarn.ApplicationMasterArguments) @bci=85, line=70 (Interpreted frame)
 - org.apache.spark.deploy.yarn.ApplicationMaster$.main(java.lang.String[]) @bci=25, line=802 (Interpreted frame)
 - org.apache.spark.deploy.yarn.ApplicationMaster.main(java.lang.String[]) @bci=4 (Interpreted frame)

到这里,已经确定是死锁,导致driver一开始就运行停滞,那么当然无法提交executor执行。
具体的死锁稍后分析,先考虑如何解决。从感性认识看,似乎只要不让kafka-client的日志也输出到kafka即可。实验后,发现果然如此:如果只输出org.apache.spark的日志就可以正常执行。

根因分析

从stack的结果看,造成死锁的是如下两个线程:

  • kafka-client内部的网络线程spark
  • 主入口线程

两个线程其实都是卡在打日志上了,观察堆栈可以发现,两个线程同时持有了同一个log对象。而这个log对象实际上是kafka-appender。而kafka-appender本质上持有kafka-client,及其内部的Metadata对象。log4j的doAppend为了保证线程安全也用synchronized修饰了:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
public
 synchronized
 void doAppend(LoggingEvent event) {
  if(closed) {
   LogLog.error("Attempted to append to closed appender named ["+name+"].");
   return;
  }
  
  if(!isAsSevereAsThreshold(event.level)) {
   return;
  }
 
  Filter f = this.headFilter;
  
  FILTER_LOOP:
  while(f != null) {
   switch(f.decide(event)) {
   case Filter.DENY: return;
   case Filter.ACCEPT: break FILTER_LOOP;
   case Filter.NEUTRAL: f = f.next;
   }
  }
  
  this.append(event); 
 }

于是事情开始了:

  • main线程尝试打日志,首先进入了synchronized的doAppend,即获取了kafka-appender的锁
  • kafka-appender内部需要调用kafka-client发送日志到kafka,最终调用到Thread 20137展示的,运行到Metadata.awaitUpdate(也是个synchronized方法),内部的wait会尝试获取metadata的锁。(详见https://github.com/apache/kaf...)
  • 但此时,kafka-producer-network-thread线程刚好进入了上文提到的打Cluster ID这个日志的这个阶段(update方法也是synchronized的),也就是说kafka-producer-network-thread线程获得了metadata对象的锁
  • kafka-producer-network-thread线程要打印日志同样执行synchronized的doAppend,即获取了kafka-appender的锁

spark通过kafka-appender指定日志输出到kafka引发的死锁问题

上图main-thread持有了log对象锁,要求获取metadata对象锁;kafka-producer-network-thread持有了metadata对象锁,要求获取log对象锁于是造成了死锁。

总结

到此这篇关于spark通过kafka-appender指定日志输出到kafka引发的死锁的文章就介绍到这了,更多相关spark指定日志输出内容请搜索服务器之家以前的文章或继续浏览下面的相关文章希望大家以后多多支持服务器之家!

原文链接:https://segmentfault.com/a/1190000022577776

延伸 · 阅读

精彩推荐