Flume allows for a -Dflume.monitoring.type=[http,ganglia] command-line argument to turn on the custom reporting mechanism [1]. I am thinking this is the way to go to emit statsd metrics.
[1] https://flume.apache.org/FlumeUserGuide.html#custom-reporting On Thu, Jul 28, 2016 at 3:07 PM, Saikat Kanjilal <[email protected]> wrote: > One other thing the agent would be shimming up instrumentation using > json/yaml etc, the aggregator would read each blob of monitoring data and > forward it to the destination > > Sent from my iPhone > > On Jul 28, 2016, at 1:06 PM, Saikat Kanjilal <[email protected]> wrote: > > I could be way off base here but I think this effort consists of two > parts, an agent in the flume channel/source/sink and a generic aggregator > module that sends data to some location (could be graphite or Cassandra or > some other nosql datastore) > > Thoughts? Sounds interesting > > Sent from my iPhone > > On Jul 28, 2016, at 12:56 PM, Ryan Phillips <[email protected]> wrote: > > That is correct. The metrics are perfectly sufficient, but I need to get > them into statsd, and then graphite. Is there a better way? > > Regards, > Ryan > > On Thu, Jul 28, 2016 at 2:54 PM, Saikat Kanjilal <[email protected]> > wrote: > >> Hey Ryan, >> Am curious what the use case for this is, in reading the wiki I see the >> following: >> >> >> https://cwiki.apache.org/confluence/display/FLUME/Monitoring+Flume >> >> Are the above metrics not sufficient or you are asking to build a tool to >> gather these metrics using statsd and potentially graphite? >> >> Sent from my iPhone >> >> On Jul 28, 2016, at 9:11 AM, Ryan Phillips <[email protected]> wrote: >> >> Greetings, >> >> Does anyone know of a Flume MonitorService that is written for StatsD? I >> am in the process of perhaps writing this module. >> >> Regards, >> Ryan >> >> >
