The statsd host name is 'labmon1001.eqiad.wmnet'.
You should use a 'prefix' for every metric that is $projectname.$hostname.<thing>. So for revisions_processed you would probably have it be ores.ores-worker-01.revisions.processed, ores.ores-workers-01.revisions.failed, etc. We can probably also add the particular model in use (enwiki/reverted, wp10 quality control, etc?) to the metric name. We can aggregate these on the other end via Graphite functions.
Things to measure:
- Counter for revisions processed
- Counter for revisions failed
- Counter for cached revisions processed
- Timing info for non-cached responses
- Timing info for cached responses