I have setup a Hadoop Cluster with Hortonworks Data Platform 2.5. I'm using 1 master and 5 slave (worker) nodes.
Every few days one (or more) of my worker nodes gets a high load and seem to restart the whole CentOS operating system automatically. After the restart the Hadoop components don't run anymore and have to be restarted manually via the Amabri management UI.
Here a screenshot of the "crashed" node (reboot after the high load value ~4 hours ago):
Here a screenshot of one of other "healthy" worker node (all other workers have similar values):
The node crashes alternate between the 5 worker nodes, the master node seems to run without problems.
What could cause this problem? Where are these high load values coming from?
This seems to be a Kernel problem, as the log file (e.g. /var/spool/abrt/vmcore-127.0.0.1-2017-06-26-12:27:34/backtrace) says something like
Version: 3.10.0-327.el7.x86_64
BUG: unable to handle kernel NULL pointer dereference at 00000000000001a0
After running a sudo yum update I had the kernel version
[root#myhost ~]# uname -r
3.10.0-514.26.2.el7.x86_64
Since the operating system updates the problem didn't occur anymore. I will observe the issue and give feedback if neccessary.
Related
I'm new to Hadoop and learning to use it by working with a small cluster where each node is an Ubuntu Server VM. The cluster consists of 1 name node and 3 data nodes with a replication factor of 3. After a power loss on the machine hosting the VMs, all files stored in the cluster were corrupted and with the blocks storing those files missing. No queries were running at the time power was lost and no files were being written to or read from the cluster.
If I shut down the VMs correctly (even without first stopping the Hadoop cluster), then the data is preserved and I don't run into any issues with missing or corrupted blocks.
The only information I've been able to find suggested setting dfs.datanode.sync.behind.writes to true, but this did not resolve the issue (killing the VMs from the host causes the same issue as a power failure). The information I found here seems to indicate this property will only have an effect when writing data to the disk.
I also tried running hdfs namenode -recover, but this did not resolve the issue. Ultimately I had to remove the data stored in the dfs.namenode.name.dir directory, rebooted each VM in the cluster to remove any Hadoop files in /tmp and reformatted the name node before copying the data back into the cluster from local file storage.
I understand that having all nodes in the cluster running on the same hardware and only 3 data nodes to go with a replication factor of 3 is not an ideal configuration, but I'd like a way to ensure that any data that is already written to disk is not corrupted by a power loss. Is there a property or other configuration I need to implement to avoid this in the future (besides separate hardware, more nodes, power backup, etc.)?
EDIT: To clarify further, the issue I'm trying to resolve is data corruption, not cluster availability. I understand I need to make changes to the overall cluster architecture to improve reliability, but I'd like a way to ensure data is not lost even in the event of a cluster-wide power failure.
I have three servers and I want to deploy Spark Standalone Cluster or Spark on Yarn Cluster on that servers.
Now I have some questions about how to allocate physical resources for a big data cluster. For example, i want to know whether i can deploy Spark Master Process and Spark Worker Process on the same node. Why?
Server Details:
CPU Cores: 24
Memory: 128GB
I need your help. Thanks.
Of course you can, just put host with Master in slaves. On my test server I have such configuration, master machine is also worker node and there is one worker-only node. Everything is ok
However be aware, that is worker will fail and cause major problem (i.e. system restart), then you will have problem, because also master will be afected.
Edit:
Some more info after question edit :) If you are using YARN (as suggested), you can use Dynamic Resource Allocation. Here are some slides about it and here article from MapR. It a very long topic how to configure memory properly for given case, I think that these resources will give you much knowledge about it
BTW. If you have already intalled Hadoop Cluster, maybe try YARN mode ;) But it's out of topic of question
I know there are other very similar questions on Stackoverflow but those either didn't get answered or didn't help me out. In contrast to those questions I put much more stack trace and log file information into this question. I hope that helps, although it made the question to become sorta long and ugly. I'm sorry.
Setup
I'm running a 9 node cluster on Amazon EC2 using m3.xlarge instances with DSE (DataStax Enterprise) version 4.6 installed. For each workload (Cassandra, Search and Analytics) 3 nodes are used. DSE 4.6 bundles Spark 1.1 and Cassandra 2.0.
Issue
The application (Spark/Shark-Shell) gets removed after ~3 minutes even if I do not run any query. Queries on small datasets run successful as long as they finish within ~3 minutes.
I would like to analyze much larger datasets. Therefore I need the application (shell) not to get removed after ~3 minutes.
Error description
On the Spark or Shark shell, after idling ~3 minutes or while executing (long-running) queries, Spark will eventually abort and give the following stack trace:
15/08/25 14:58:09 ERROR cluster.SparkDeploySchedulerBackend: Application has been killed. Reason: Master removed our application: FAILED
org.apache.spark.SparkException: Job aborted due to stage failure: Master removed our application: FAILED
at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1185)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1174)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1173)
at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59)
at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47)
at org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1173)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:688)
at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:688)
at scala.Option.foreach(Option.scala:236)
at org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:688)
at org.apache.spark.scheduler.DAGSchedulerEventProcessActor$$anonfun$receive$2.applyOrElse(DAGScheduler.scala:1391)
at akka.actor.ActorCell.receiveMessage(ActorCell.scala:498)
at akka.actor.ActorCell.invoke(ActorCell.scala:456)
at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:237)
at akka.dispatch.Mailbox.run(Mailbox.scala:219)
at akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinTask.exec(AbstractDispatcher.scala:386)
at scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260)
at scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339)
at scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
at scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)
FAILED: Execution Error, return code -101 from shark.execution.SparkTask
This is not very helpful (to me), that's why I'm going to show you more log file information.
Error Details / Log Files
Master
From the master.log I think the interesing parts are
INFO 2015-08-25 09:19:59 org.apache.spark.deploy.master.DseSparkMaster: akka.tcp://sparkWorker#172.31.46.48:46715 got disassociated, removing it.
INFO 2015-08-25 09:19:59 org.apache.spark.deploy.master.DseSparkMaster: akka.tcp://sparkWorker#172.31.33.35:42136 got disassociated, removing it.
and
ERROR 2015-08-25 09:21:01 org.apache.spark.deploy.master.DseSparkMaster: Application Shark::ip-172-31-46-49 with ID app-20150825091745-0007 failed 10 times, removing it
INFO 2015-08-25 09:21:01 org.apache.spark.deploy.master.DseSparkMaster: Removing app app-20150825091745-0007
Why do the worker nodes get disassociated?
In case you need to see it, I attached the master's executor (ID 1) stdout as well. The executors stderr is empty. However, I think it shows nothing useful to tackle the issue.
On the Spark Master UI I verified to see all worker nodes to be ALIVE. The second screenshot shows the application details.
There is one executor spawned on the master instance while executors on the two worker nodes get respawned until the whole application is removed. Is that okay or does it indicate some issue? I think it might be related to the "(it) failed 10 times" error message from above.
Worker logs
Furthermore I can show you logs of the two Spark worker nodes. I removed most of the class path arguments to shorten the logs. Let me know if you need to see it. As each worker node spawns multiple executors I attached links to some (not all) executor stdout and stderr dumps. Dumps of the remaining executors look basically the same.
Worker I
worker.log
Executor (ID 10) stdout
Executor (ID 10) stderr
Worker II
worker.log
Executor (ID 3) stdout
Executor (ID 3) stderr
The executor dumps seem to indicate some issue with permission and/or timeout. But from the dumps I can't figure out any details.
Attempts
As mentioned above, there are some similar questions but none of those got answered or it didn't help me to solve the issue. Anyway, things I tried and verified are:
Opened port 2552. Nothing changes.
Increased spark.akka.askTimeout which results in the Spark/Shark app to live longer but eventually it still gets removed.
Ran the Spark shell locally with spark.master=local[4]. On the one hand this allowed me to run queries longer than ~3 minutes successfully, on the other hand it obviously doesn't take advantage of the distributed environment.
Summary
To sum up, one could say that the timeouts and the fact long-running queries are successfully executed in local mode all indicate some misconfiguration. Though I cannot be sure and I don't know how to fix it.
Any help would be very much appreciated.
Edit: Two of the Analytics and two of the Solr nodes were added after the initial setup of the cluster. Just in case that matters.
Edit (2): I was able to work around the issue described above by replacing the Analytics nodes with three freshly installed Analytics nodes. I can now run queries on much larger datasets without the shell being removed. I intend not to put this as an answer to the question as it is still unclear what is wrong with the three original Analytics nodes. However, as it is a cluster for testing purposes, it was okay to simply replace the nodes (after replacing the nodes I performed a nodetool rebuild -- Cassandra on each of the new nodes to recover their data from the Cassandra datacenter).
As mentioned in the attempts, the root cause is a timeout between the master node, and one or more workers.
Another thing to try: Verify that all workers are reachable by hostname from the master, either via dns or an entry in the /etc/hosts file.
In my case, the problem was that the cluster was running in an AWS subnet without DNS. The cluster grew over time by spinning up a node, the adding the node to the cluster. When the master was built, only a subset of the addresses in the cluster was known, and only that subset was added to the /etc/hosts file.
When dse spark was run from a "new" node, then communication from the master using the worker's hostname failed and the master killed the job.
I have installed the hadoop 2.2 with four machines. They are:
namenodes: master1,master2
datanodes: slave1,slave2
The master1 is installed on my notebook, and I want to close the master1 when I sleep. And
when the master1 is in the active state, the master2 is in the standby state. When I close my notebook, will the hadoop cluster automatically change the active namenode to the master2?
I don't know if I understand the meaning of hadoop v2's multiple namenodes. Does the feature fit my situation described above? Thanks.
You need to use the High Availability feature for the NameNodes. When it detects the Primary NameNode going offline, the Secondary will automatically take over.
There will be a brief hiccup in the operation of the cluster, however, as the Secondary node will delay in responding to block location requests for a short amount of time (30 to 60 seconds usually), giving all Data nodes enough time to point to the new NN.
I have to explicitly boot all the processor nodes every time when my master system restarts
i.e
mdpboot -n 3
P.S
I am implementing beowulf cluster
Try updating to a new version of MPICH. MPD is quite old and hasn't been supported for some time. The newer versions use Hydra which doesn't require booting.