topic Re: How to configure Field Perf Benchmark App in Monitoring Splunk

How to configure Field Perf Benchmark App

Blu3fish — Sat, 17 Jul 2010 03:30:53 GMT

What actually needs to be configured within Perf Test's tasks.conf file.

Also do the following global values need to be manually defined within benchmark.py? Starting on line 26:

config = None
resultsDir = None
resultsStream = None
logStream = None
splunkHome = None
splunkDb = None
splunkVersion = None
splunkCmd = None
timestamp = None
timestampForFilenames = None
authArg = None

This could probably be squeezed under this thread: http://answers.splunk.com/questions/4671/how-to-use-field-perf-benchmark-app

Re: How to configure Field Perf Benchmark App

gfriedmann — Sat, 05 Feb 2011 09:36:12 GMT

You should specify a directory with data to load and use for benchmarking. Then you run the benchmark script. Wait for it to complete, then check out the web interface for pretty results.

Repeat the benchmark for fun. Maybe the tasks will need new names (i'm not sure yet).

Make some changes to the system, and run it again. I think this is more of a system benchmark independant of your actual individual dataset.

Re: How to configure Field Perf Benchmark App

V_at_Splunk — Sat, 05 Feb 2011 11:35:55 GMT

What needs to be in tasks.conf:

the dataset (specifically, absolute path to a file) that you want to index, to see how long the indexing takes; this is an index task
if you want to time some searches on the resulting index (that was generated by the index task above), specify search tasks

This is not a system benchmark. For that, you could try adapting some of the UNIX App (or Windows App) scripts. The idea with Field Perf Benchmark App, was an OS-agnostic perf measurement tool.

This app is, indeed, all about your individual dataset (and, of course, your individual hardware).

And finally, no, you should not need to modify anything in benchmark.py script.

Re: How to configure Field Perf Benchmark App

V_at_Splunk — Sat, 05 Feb 2011 11:37:11 GMT

No, the tasks will not need new names, for when you run it again. The reports know enough to differentiate between successive runs.