perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Zhi-Wei Lu writes:

 

After I put

                    "127.0.0.1/32",

                    "::1/128"

 

In the limits.conf file in the “allow” hosts,

pscheduler troubleshoot

worked just fine for the “localhost” trouble shooting.  However, pscheduler still fail between a few servers.

 

That description sounds like you’re using a custom limit configuration.  The one we ship with the toolkit (see https://github.com/perfsonar/toolkit/blob/master/etc/default_service_configs/pscheduler_limits.conf) has an identifier called “local-interfaces” which is folded into the “friendlies” classifier which in turn is allowed to do anything.  That should allow any task request coming from the local machine to be granted.  If you’d like to send me your limit configuration off-list, I’ll have a look at it.

 

 

On pb.noc.ucdavis.edu, I had problem to talk to pscheduler on p0.noc.ucdavis.edu

 …

  Checking path MTU... Unsafe or unknown: Error: Process took too long to run.

  Checking for pScheduler... Failed.

Request timed out

 

I wonder what it prevents pscheduler to talk to p0.noc.ucdavis.edu from pb.noc.ucdavis.edu.

 

I was able to reach pScheduler on both hosts from outside of UC Davis and ran some tests between them.  pScheduler is working fine on both and ICMP echo traffic and the UDP packets to make traceroute function work in both directions.  HTTPS traffic, which is how pSchedulers communicate with each other, does not, suggesting that something in the network between them is blocking it.

 

--Mark