perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Tao Zhang writes:

 

All our perfSONAR boxes have been upgraded to 5.0, however, the archive shows empty. 

…

I checked the pscheduler log. The log filled with something like this:

Apr 21 12:27:31 lat-sask1pfs1 journal: archiver WARNING  120686191: Failed to archive https://lat-sask1pfs1.network.canarie.ca/pscheduler/tasks/b4843fdf-6d63-4a72-a29e-770134ccf853/runs/b7ed7962-377d-4842-90da-952c3e3670ea to esmond: Archiver permanently abandoned registering test after 1 attempt(s): 405: Invalid JSON returned

…

Apr 21 12:27:34 lat-sask1pfs1 journal: archiver WARNING  <title>503 Service Unavailable</title>

 

5.0 replaced Esmond with LogStash and OpenSearch.  There is a backward-compatibility layer for Esmond, but it is read-only.  If your tasks are archiving to the local system, they will need to use the HTTP archiver and post it to LogStash instead.  These two sections of the documentation should get you there:

 

• https://docs.perfsonar.net/multi_ma_install.html#authenticating-measurement-hosts
• https://docs.perfsonar.net/multi_ma_install.html#configuring-measurement-hosts

 

 

Apr 21 12:27:30 lat-sask1pfs1 journal: runner ERROR      File "/usr/libexec/pscheduler/classes/tool/traceroute/run", line 239, in <module>

Apr 21 12:27:30 lat-sask1pfs1 journal: runner ERROR        run_timeout += send_wait * hops

Apr 21 12:27:30 lat-sask1pfs1 journal: runner ERROR    TypeError: unsupported operand type(s) for +=: 'datetime.timedelta' and 'str'

 

This is a bug in pScheduler that appears to have been introduced in the last release.  I’ve already corrected it and added it to the pipeline for the next bugfix release, which should happen sometime during the week of May 1.

 

Based on what I could pull out of the API on your system, these are the tasks that are failing:

 

2023-04-21T14:27:24-06:00 - 2023-04-21T14:30:32-06:00 (Failed)

trace --dest lat-gp-mn.nbren.ca --length 40 --probe-type udp --source lat-sask1pfs1.network.canarie.ca --source-node lat-sask1pfs1.network.canarie.ca --hops 30 --wait PT30S --sendwait PT5S (Run with tool 'traceroute')

https://lat-sask1pfs1.network.canarie.ca/pscheduler/tasks/1aa55201-d3dd-49b4-bda6-99524115462e/runs/0c7a63c4-70bc-4eeb-8181-dfcb4195d148

 

As a stopgap, removing the ‘sendwait’ option from your trace tasks won’t trip the faulty code.  Note that this will alter the character of the measurement.

 

--Mark