accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Billie Rinaldi <bil...@apache.org>
Subject Re: Programtically invoking a Map/Reduce job
Date Thu, 17 Jan 2013 19:57:18 GMT
On Thu, Jan 17, 2013 at 11:16 AM, Mike Hugo <mike@piragua.com> wrote:

> Thanks Billie!
>
> Setting "mapred.job.tracker" and "fs.default.name" in the conf has gotten
> me further.
>
>          job.getConfiguration().set("mapred.job.tracker",
> "server_name_here:8021");
>         job.getConfiguration().set("fs.default.name",
> "hdfs://server_name_here:8020");
>
> What's interesting now is that the job can't find Accumulo classes - when
> I run the job now, I get
>
> 2013-01-17 12:59:25,278 [main] INFO  mapred.JobClient  - Task Id :
> attempt_201301171102_0012_m_000000_1, Status : FAILED
> java.lang.RuntimeException: java.lang.ClassNotFoundException:
> org.apache.accumulo.core.client.mapreduce.AccumuloRowInputFormat
>
> Is there a way to inform the job (via the Job API, on a separate machine
> not running hadoop) about extra libs to include on the classpath of the job?
>

You normally inform a job about jars it needs by specifying "-libjars
comma,separated,jar,list" on the command line.  In this case, you need to
put those two strings "-libjars" and "jar,list" in the String[] args passed
to ToolRunner.run:
ToolRunner.run(CachedConfiguration.getInstance(), new ...(), args)

The accumulo-core jar probably isn't the only one you'll need.

Billie


>
> Thanks
>
> Mike
>
>
>
> On Wed, Jan 16, 2013 at 3:11 PM, Billie Rinaldi <billie@apache.org> wrote:
>
>> Your job is running in "local" mode (Running job: job_local_0001).  This
>> basically means that the hadoop configuration is not present on the
>> classpath of the java client kicking off the job.  If you weren't planning
>> to have the hadoop config on that machine, you might be able to get away
>> with setting "mapred.job.tracker" and probably also "fs.default.name" on
>> the Configuration object.
>>
>> Billie
>>
>>
>>
>> On Wed, Jan 16, 2013 at 12:07 PM, Mike Hugo <mike@piragua.com> wrote:
>>
>>> Cool, thanks for the feedback John, the examples have been helpful in
>>> getting up and running!
>>>
>>> Perhaps I'm not doing something quite right.  When I jar up my jobs and
>>> deploy the jar to the server and run it via the tool.sh command on the
>>> cluster, I see the job running in the jobtracker (servername:50030) and it
>>> runs as I would expect.
>>>
>>> 13/01/16 14:39:53 INFO mapred.JobClient: Running job:
>>> job_201301161326_0006
>>> 13/01/16 14:39:54 INFO mapred.JobClient:  map 0% reduce 0%
>>> 13/01/16 14:41:29 INFO mapred.JobClient:  map 50% reduce 0%
>>> 13/01/16 14:41:35 INFO mapred.JobClient:  map 100% reduce 0%
>>> 13/01/16 14:41:40 INFO mapred.JobClient: Job complete:
>>> job_201301161326_0006
>>> 13/01/16 14:41:40 INFO mapred.JobClient: Counters: 18
>>> 13/01/16 14:41:40 INFO mapred.JobClient:   Job Counters
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     SLOTS_MILLIS_MAPS=180309
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Total time spent by all
>>> reduces waiting after reserving slots (ms)=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Total time spent by all
>>> maps waiting after reserving slots (ms)=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Rack-local map tasks=2
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Launched map tasks=2
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     SLOTS_MILLIS_REDUCES=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:   File Output Format Counters
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Bytes Written=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:   FileSystemCounters
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     HDFS_BYTES_READ=248
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     FILE_BYTES_WRITTEN=60214
>>> 13/01/16 14:41:40 INFO mapred.JobClient:   File Input Format Counters
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Bytes Read=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:   Map-Reduce Framework
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Map input records=1036434
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Physical memory (bytes)
>>> snapshot=373760000
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Spilled Records=0
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     CPU time spent (ms)=24410
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Total committed heap usage
>>> (bytes)=168394752
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Virtual memory (bytes)
>>> snapshot=2124627968
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     Map output records=2462684
>>> 13/01/16 14:41:40 INFO mapred.JobClient:     SPLIT_RAW_BYTES=248
>>>
>>>
>>>
>>> When I kick off a job via a java client running on a different host, the
>>> job seems to run (I can see things being scanned and ingested) but I don't
>>> see anything via the jobtracker UI on the server.  Is that normal?  Or do I
>>> have something mis-configured?
>>>
>>>
>>>
>>> Here's how I'm starting things from the client:
>>>
>>>     @Override
>>>     public int run(String[] strings) throws Exception {
>>>         Job job = new Job(getConf(), getClass().getSimpleName());
>>>         job.setJarByClass(getClass());
>>>         job.setMapperClass(MyMapper.class);
>>>
>>>         job.setInputFormatClass(AccumuloRowInputFormat.class);
>>>
>>> AccumuloRowInputFormat.setZooKeeperInstance(job.getConfiguration(),
>>> instanceName, zookeepers);
>>>
>>>         AccumuloRowInputFormat.setInputInfo(job.getConfiguration(),
>>>                 username,
>>>                 password.getBytes(),
>>>                 "...",
>>>                 new Authorizations());
>>>
>>>         job.setNumReduceTasks(0);
>>>
>>>         job.setOutputFormatClass(AccumuloOutputFormat.class);
>>>         job.setOutputKeyClass(Key.class);
>>>         job.setOutputValueClass(Mutation.class);
>>>
>>>         boolean createTables = true;
>>>         String defaultTable = "...";
>>>         AccumuloOutputFormat.setOutputInfo(job.getConfiguration(),
>>>                 username,
>>>                 password.getBytes(),
>>>                 createTables,
>>>                 defaultTable);
>>>
>>>
>>> AccumuloOutputFormat.setZooKeeperInstance(job.getConfiguration(),
>>> instanceName, zookeepers);
>>>
>>>         job.waitForCompletion(true);
>>>
>>>         return job.isSuccessful() ? 0 : 1;
>>>     }
>>>
>>>     public static void main(String args[]) throws Exception {
>>>         int res = ToolRunner.run(CachedConfiguration.getInstance(), new
>>> ...(), args);
>>>         System.exit(res);
>>>     }
>>>
>>>
>>>
>>> Here's the output when I run it via the client application:
>>>
>>>
>>> 2013-01-16 13:55:57,645 [main-SendThread()] INFO  zookeeper.ClientCnxn
>>>  - Opening socket connection to server accumulo/10.1.10.160:2181
>>> 2013-01-16 13:55:57,660 [main-SendThread(accumulo:2181)] INFO
>>>  zookeeper.ClientCnxn  - Socket connection established to accumulo/
>>> 10.1.10.160:2181, initiating session
>>> 2013-01-16 13:55:57,671 [main-SendThread(accumulo:2181)] INFO
>>>  zookeeper.ClientCnxn  - Session establishment complete on server accumulo/
>>> 10.1.10.160:2181, sessionid = 0x13c449cfe010434, negotiated timeout =
>>> 30000
>>> 2013-01-16 13:55:58,379 [main] INFO  mapred.JobClient  - Running job:
>>> job_local_0001
>>> 2013-01-16 13:55:58,447 [Thread-16] INFO  mapred.Task  -  Using
>>> ResourceCalculatorPlugin : null
>>> 2013-01-16 13:55:59,383 [main] INFO  mapred.JobClient  -  map 0% reduce
>>> 0%
>>> 2013-01-16 13:56:04,458 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:07,459 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:10,461 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:13,462 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:16,463 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:19,465 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:21,783 [Thread-16] INFO  mapred.Task  -
>>> Task:attempt_local_0001_m_000000_0 is done. And is in the process of
>>> commiting
>>> 2013-01-16 13:56:21,783 [Thread-16] INFO  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:21,784 [Thread-16] INFO  mapred.Task  - Task
>>> 'attempt_local_0001_m_000000_0' done.
>>> 2013-01-16 13:56:21,786 [Thread-16] INFO  mapred.Task  -  Using
>>> ResourceCalculatorPlugin : null
>>> 2013-01-16 13:56:22,423 [main] INFO  mapred.JobClient  -  map 100%
>>> reduce 0%
>>> 2013-01-16 13:56:27,788 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:28,440 [main] INFO  mapred.JobClient  -  map 50% reduce
>>> 0%
>>> 2013-01-16 13:56:30,790 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:33,791 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:36,792 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:39,793 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:42,794 [communication thread] INFO
>>>  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:45,779 [Thread-16] INFO  mapred.Task  -
>>> Task:attempt_local_0001_m_000001_0 is done. And is in the process of
>>> commiting
>>> 2013-01-16 13:56:45,780 [Thread-16] INFO  mapred.LocalJobRunner  -
>>> 2013-01-16 13:56:45,781 [Thread-16] INFO  mapred.Task  - Task
>>> 'attempt_local_0001_m_000001_0' done.
>>> 2013-01-16 13:56:45,782 [Thread-16] WARN  mapred.FileOutputCommitter  -
>>> Output path is null in cleanup
>>> 2013-01-16 13:56:46,462 [main] INFO  mapred.JobClient  -  map 100%
>>> reduce 0%
>>> 2013-01-16 13:56:46,462 [main] INFO  mapred.JobClient  - Job complete:
>>> job_local_0001
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  - Counters: 7
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -
>>> FileSystemCounters
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -
>>> FILE_BYTES_READ=1257
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -
>>> FILE_BYTES_WRITTEN=106136
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -   Map-Reduce
>>> Framework
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -     Map input
>>> records=1036434
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -     Spilled
>>> Records=0
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -     Total
>>> committed heap usage (bytes)=259915776
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -     Map output
>>> records=2462684
>>> 2013-01-16 13:56:46,463 [main] INFO  mapred.JobClient  -
>>> SPLIT_RAW_BYTES=240
>>>
>>>
>>>
>>> On Wed, Jan 16, 2013 at 11:20 AM, John Vines <vines@apache.org> wrote:
>>>
>>>> The code examples we have scripted simply do the necessary setup for
>>>> creating a mapreduce job and kicking it off. If you check out the code for
>>>> them in
>>>> src/examples/simple/src/main/java/org/apache/accumulo/examples/simple/mapreduce/
>>>> you can see what we're doing in Java to kick off jobs.
>>>>
>>>> The short explanation is, just like any other MapReduce job, we're
>>>> setting up a Job, configuring the AccumuloInput and/or OutputFormats, and
>>>> sending them off like any other MapReduce job.
>>>>
>>>> John
>>>>
>>>>
>>>> On Wed, Jan 16, 2013 at 12:11 PM, Mike Hugo <mike@piragua.com> wrote:
>>>>
>>>>> I'm writing a client program that uses the BatchWriter and
>>>>> BatchScanner for inserting and querying data, but occasionally it also
>>>>> needs to be able to kick of a Map/Reduce job on a remote accumulo cluster.
>>>>>  The Map/Reduce examples that ship with Accumulo look like they are meant
>>>>> to be invoked via the command line.  Does anyone have an example of how
to
>>>>> kick something off via a java client running on a separate server?  Any
>>>>> best practices to share?
>>>>> Thanks,
>>>>> Mike
>>>>>
>>>>
>>>>
>>>
>>
>

Mime
View raw message