On Wed, May 16, 2018 at 1:22 AM, Chamarthy, Mahati <mahati.chamarthy@xxxxxxxxx> wrote: > +ceph-devel > > From: Chamarthy, Mahati > Sent: Wednesday, May 16, 2018 1:46 PM > To: > Cc: dillaman@xxxxxxxxxx > Subject: rbd tracing efforts using zipkin > > I had recently setup zipkin and was able to pass along ceph lttng traces for visualization. > > The visualization is pretty helpful, however due to the large number of I/O requests, there is a need for sampling. Has there been work on this side? Or how is this handled currently when one wants to establish the numbers for profiling? The zipkin stuff is pretty much abandoned after it was merged into the codebase, so I don't know of any efforts to directly improve it. There are some discussions in this mailing list about integrating w/ OpenTracing as well, but I'm semi-out-of-the-loop on it right now. Adding optional sampling to the trace instantiation within librbd makes sense to me. > A better question is perhaps, what is the best way to make a performance comparison on rbd I/O path before/after a change to librbd/librados code? fio benchmarks and perhaps sar (to watch CPU utilization)? > Thanks, > Mahati > > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- Jason -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
