+1 The underlying impl which is IIRC Codehale metrics used sun.misc.Unsafe pre 3.2 fixed in [1]. It should be updated once 3.2 is updated. I am not certain how Metrics 3.1 behaves when run on a non Sun JDK. I think it falls back to a slower impl of Stripe64 that doesn't use Unsafe internally.
I don't think this should block adding metrics and instrumenting operations with timers/counters etc. As more metrics are added, I think care needs to be taken to ensure the reporting mechanisms don't become resource hogs. IIRC some reporters (eg JMX, Console) can become expensive with 1000s of metrics. Ganglia, JSON, ES etc reporters are all fine with large numbers of metrics. Best Regards Ian 1 https://github.com/dropwizard/metrics/issues/539 On 31 August 2016 at 09:33, Chetan Mehrotra <[email protected]> wrote: > +1. This would be useful > > Some ideas on what metrics need to be collected at mentioned in SLING-5410. > Chetan Mehrotra > > > On Wed, Aug 31, 2016 at 1:56 PM, Bertrand Delacretaz > <[email protected]> wrote: > > Hi, > > > > Shall we generally add metrics [1] to our core? > > > > I have a student at $work that needs to do a small project, I thought > > they could contribute to that, including creating a demo setup with > > some visualization tool. > > > > We'd need to define guidelines for which metrics to add where but > > that's a good thing anyway. > > > > WDYT? > > > > -Bertrand > > > > [1] https://sling.apache.org/documentation/bundles/metrics.html >
