perfSONAR User Q&A and Other Discussion

[Brian Candler <>]

On 05/09/2019 09:45, Brian Candler wrote:

As for point (1), even if priority is only supposed to be used for conflicting tests, I still don't understand yet why the priority was being set differently for my inbound and outbound tests. 

I think I found it: priority.script in /etc/pscheduler/limits.conf.  It sets priority 5 if "requester" is in "friendlies", which in turn is "local-interfaces".

I presume that the "requester" is the lead participant; and even though I typed the command on the local host, it chooses the --source of the test as the lead participant.

Note: the reference documentation for the throughput test plugin doesn't say explicitly that the --source is also the lead participant.  However it says here:

For many tests run by perfSONAR, a --source switch which specifies where the test should originate and is also the lead participant:

So it's true for "many tests".

I'd say this leads to surprising behaviour for the end user.  It means that a throughput test which consumes my outbound bandwidth always gets a higher priority than a test which consumes my inbound bandwidth.  This is regardless of whether the test was created by me, or by some random remote user.

What I would *expect* is that any test that I have created (inbound or outbound) would get priority over any test submitted by a random remote user (inbound or outbound).

Anyway, that's a different issue altogether, and not really a major problem, although it might be nice to document it clearly as a consequence of the pscheduler design.

Regards,

Brian Candler.