Network and Transfer Metrics WG - perfSONAR operations

Europe/Zurich
28/R-015 (CERN)

28/R-015

CERN

15
Show room on map
Marian Babik (CERN), Shawn McKee (University of Michigan ATLAS Group)
Description

The meeting date/time is a result of a fixed schedule agreed in the last meeting (https://indico.cern.ch/event/354593/).

The purpose of the meeting will be mainly to discuss news from the perfSONAR workshop, second broadcast and follow up, proposed changes to the infrastructure monitoring and mesh configurations as well as LHCOPN/LHCONE situation. 

Details on the Network and Transfer Metrics WG are available at our Twiki.  

Slides

Network and Transfer Metrics WG - perfSONAR operations (28 Jan 2015)
Chaired by: Dr. McKee, Shawn; Babik, Marian

Participants:  Alessandro De Salvo, Bruno Hoeft, Duncan Rand, Enrico Mazzoni, Frederique Chollet, Hung-Te Lee, Ian Gable, Stefan Stancu, Shawn McKee, Marian Babik

Minutes:

Current status and tools were presented followed by demo of the configuration interface, infrastructure monitoring and short introduction to the datastore. 

Alessandro commented on hosts at Rome "down" (from blocked ping) but all tests OK.  

Marian: Host up/down states can be ignored as they're usually due to blocked ICMP, it's important that all tests are OK. We can try to change the current ping test to make all hosts UP.

Bruno asked if host can participate in multiple meshes.

Marian: Yes, hosts can participate in any number of meshes and different meshes can be maintained by different people depending on the scope of particular mesh. Thanks to the auto-url configuration any given sonar will receive all the tests that are relevant for it. 

Duncan would like to test the configuration interface for the IPv6. To be followed up.

It was commented that while we do have auto-updates of the package, we still need to ask sites to reboot in order for the updates to take effect. 

It was agreed that we'll keep inter-cloud tests disabled and that we'll gradually introduce a full mesh latency tests while at the same time temporarily suspend full mesh bandwidth tests (to reduce the load on OPN hosts). Full mesh latency tests will be started between top-k sites (k will be based on the total amount of online storage reported in REBUS). We'll keep full mesh traceroute tests in place - note that those were migrated to bandwidth hosts (since they require bwctl in certain cases).

There will be a second broadcast to announce availability of perfSONAR infrastructure monitoring (http://pfomd.grid.iu.edu) and remind sites to update to 3.4.1. We plan to follow it up with a campaign to bring all sonars to the correct configuration, starting with OPN and LHCONE 

Finally, Shawn gave a short summary of the perfSONAR workshop.
 

There are minutes attached to this event. Show them.
The agenda of this meeting is empty