perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Szymon Trocha writes:

I was able to replicate the issue so I opened a report in Github https://github.com/perfsonar/pscheduler/issues/1423

That helps and I was able to find the problem.  This one’s fun.

When pScheduler’s ping plugin was first written, ping accepted 20.0 as a value for the deadline switch.  At some point, they changed how the value was parsed changed and 20.0 became invalid, causing ping to exit with a status of 1.  The systems where I tested this were running the older version and Thomas and Szymon’s had the newer one.  One quirk of ping is that it exits 1 whether there was a genuine problem or if the remote host didn’t respond.  The plugin was taking an exit status of 1 as the latter and returning 100% packet loss, which is a reasonable thing to do if the arguments are all valid.

I’ll add code to the next release to make the plugin able to tell the difference between no answer and actual errors.

This may also be related to https://github.com/perfsonar/pscheduler/issues/1422

I wrote that yesterday while experimenting with this problem.  It’s unrelated, and would have caused test failures with a “program took too long to run” error.  That’ll be fixed in the next release, too.

--Mark