perfSONAR User Q&A and Other Discussion

[Jianan Wang <>]

Thanks Mark for your detailed explanation! Looks like using the json format query is pretty handy in this situation. Will try it out then:)

Jianan.

On Thu, May 24, 2018 at 2:31 PM, Mark Feit <> wrote:

Jianan Wang writes:

    Recently I'm thinking about collecting the tasks scheduled by perfsonar in
    real time and use that information programmatically to do some network traffic
    job splitting related experiment. Therefore, just want to understanding if
    there is any support from this perspective and what are the API or query
    methods I could use to achieve it.

If I understand what you're asking correctly, you'd like to stream measurement results out of perfSONAR into your own system.  We can do that.

pScheduler, which is what's doing the measurements, has a called archivers that you can use to reliably send results to destinations other than the Esmond measurement archive that's been standard for a long time.  There are plugins for Esmond, HTTP POST, RabbitMQ and syslog shipped with the software, which should get you to a lot of the places you'll need to go.  The hitch is that right now, MeshConfig is only capable of generating tasks that archive to Esmond.  perfSONAR 4.1, which is about to go into beta, replaces MeshConfig with pSConfig, which lets you use all of pScheduler's features, including all of the archivers.

The good news is that if you want to do some experimentation before we get 4.1 released, there's a way around that limitation.  You can either run single tasks that specify other archivers or install a file on one of your systems that will force pScheduler to archive all measurements to an archiver, even those that didn't specify it (this is called "global archiving").   Documentation on how to use them is available at http://docs.perfsonar.net/release_candidates/4.1b1/pscheduler_ref_archivers.html.  Note that although this is beta documentation, everything described is available in the current production code and has been for quite awhile.

--Mark

--

Jianan Wang