Error while trying to run a MapReduce job on FIWARE-Cosmos using Tidoop REST API

Question

I am following this guide on Github and I am not able run the example mapreduced job mentioned in Step 5.

I am aware that this file no longer exists:

/usr/lib/hadoop-mapreduce/hadoop-mapreduce-examples.jar

And I am aware that the same file can now be found here:

/usr/lib/hadoop-0.20/hadoop-examples-0.20.2-cdh3u6.jar

So I form my call as below:

curl -v -X POST "http://computing.cosmos.lab.fiware.org:12000/tidoop/v1/user/$user/jobs" -d '{"jar":"/usr/lib/hadoop-0.20/hadoop-examples-0.20.2-cdh3u6.jar","class_name":"WordCount","lib_jars":"/usr/lib/hadoop-0.20/hadoop-examples-0.20.2-cdh3u6.jar","input":"testdir","output":"testoutput"}' -H "Content-Type: application/json" -H "X-Auth-Token: $TOKEN"

The input directory exists in my hdfs user space and there is a file called testdata.txt inside it. The testoutput folder does not exist in my hdfs user space since I know it creates problems.

When I execute this curl command, the error I get is {"success":"false","error":1} which is not very descriptive. Is there something I am missing here?

Answer 1

This has been just tested with my user frb and a valid token for that user:

$ curl -X POST "http://computing.cosmos.lab.fiware.org:12000/tidoop/v1/user/frb/jobs" -d '{"jar":"/usr/lib/hadoop-mapreduce/hadoop-mapreduce-examples.jar","class_name":"wordcount","lib_jars":"/usr/lib/hadoop-mapreduce/hadoop-mapreduce-examples.jar","input":"testdir","output":"outputtest"}' -H "Content-Type: application/json" -H "X-Auth-Token: xxxxxxxxxxxxxxxxxxx"
{"success":"true","job_id": "job_1460639183882_0011"}

Please observe the fat jar with the MapReduce examples in the "new" cluster (computing.cosmos.lab.fiware.org) is at /usr/lib/hadoop-mapreduce/hadoop-mapreduce-examples.jar, as detailed in the documentation. /usr/lib/hadoop-0.20/hadoop-examples-0.20.2-cdh3u6.jar was the fat jar in the "old" cluster (cosmos.lab.fiware.org).

EDIT 1

Finally, the user had no account in the "new" pair of clusters of Cosmos in FIWARE LAB (storage.cosmos.lab.fiware.org and computing.cosmos.lab.fiware.org), where Tidoop runs, but in another "old" cluster (cosmos.lab.fiwre.org). Thus, the issue was fixed by simply provisioning an account in the "new" ones.