- metric - the tool or benchmark that gathered data
- metric type - transactions, throughput, etc
- iteration - a uniqe set of benchmark parameters
- sample - a re-run of an iteration for greater statistical confidence
- cstype - a worker, master, client, or server. initially was short for client server type. but with workers and masters the name is ill fitting
- csid - the id of the cstype, worker-1 for example
- package - like a numanode
- type - the type of cpu, user, sys, etc
Metrics are collected over a time range.
Crucible only grabs metrics that are the most granular / "broken out" metric that the tools provide. IE if a tool reports the average cpu usage, crucible doesn't collect it. Crucible Aggregates in Elastic Search.
- Select your run id
- select the time range within the run id
- select the sources of data that you wish, mpstat, uperf for example
- run the query via
crucible get
Run id is found in rickshaw-run.json
inside /var/lib/crucible/run/
crucible get result --run <RUN-ID>
, will list all metrics available, by source and typecrucible get metric --run <RUN-ID> --begin <start> --end <final_time> --source mpstat --type Busy-CPU
will return a value, but also provides breakouts
, which can be thought of as metadata.
-
crucible get metric --run <RUN-ID> --begin <start> --end <final_time> --source mpstat --type Busy-CPU --breakout cstype,csid
-
crucible get metric --run <RUN-ID> --begin <start> --end <final_time> --source mpstat --type Busy-CPU --breakout cstype=worker,csid=1,package
-
use
--resolution=90
to get the desired number of samples from the run. just like a picture you can only enhance so far before it doesn't get any clearer. try a resolution that matches the duration of a test -
crucible get metric --run <RUN-ID> --begin <start> --end <final_time> --source mpstat --type Busy-CPU --breakout cstype,csid --resolution 60