I am attempting to run the SLS Load simulator to test different yarn schedulers.
When I try to start the sls simulator, I get the following parsing error.
Exception in thread "main" java.lang.NullPointerException
at org.apache.hadoop.yarn.sls.utils.SLSUtils.parseNodesFromRumenTrace(SLSUtils.java:69)
at org.apache.hadoop.yarn.sls.SLSRunner.startNM(SLSRunner.java:189)
at org.apache.hadoop.yarn.sls.SLSRunner.start(SLSRunner.java:143)
at org.apache.hadoop.yarn.sls.SLSRunner.main(SLSRunner.java:528)
If I try to convert this rumen to sls then I get a different error.
$:~/hadoop-2.6.0/share/hadoop/tools/sls$ bin/rumen2sls.sh --rumen-file=job-trace.json --output-dir=sls_out
Exception in thread "main" java.lang.ClassCastException: java.lang.Integer cannot be cast to java.lang.Long
at org.apache.hadoop.yarn.sls.RumenToSLSConverter.createSLSJob(RumenToSLSConverter.java:170)
at org.apache.hadoop.yarn.sls.RumenToSLSConverter.generateSLSLoadFile(RumenToSLSConverter.java:132)
at org.apache.hadoop.yarn.sls.RumenToSLSConverter.main(RumenToSLSConverter.java:116)
It seems that I have some corrupt data, or unreadable data for the Simulator. Is there any easy way to hunt down the unreadable data?
Sometimes, when folding the file, I will randomly get an output that works. Probably because it randomly ended up avoiding the unreadable jobs, however I would like to run the simulator on my entire set of jobs.