WLCG HTTP Deployment TF Meeting 11^th Nov.

Present

Marian
Cédric
Oliver
Christophe
Dave
Duncan Rand
Enrico
Georgios
Luca
Xavier

Minutes

Review of previous actions

All topology will be moved to VO feed, which means this is now part of a larger discussion about VO feed for all services. This will be sorted on a longer timescale so we stick with the current setup (xls from LHCb, AGIS query for Atlas) for now.

Firewall has been opened.

Marian finalising puppetisation. He will now move Alice onto preprod, followed by LHCb and Atlas. The effect will be that the results of other tests will also appear in the interface, and SAM3 will be available for visualisation. This move will require some downtime.

Experiment statement for site twiki - "not critical but encouraged" is the general message, something official will have to wait.

Volunteers

Discussion on IPv6. Conclusion was that it's best to leave IPv6-only stuff to the IPv6 Working Group (thanks to Oxford for volunteering), otherwise a dedicated ETF instance would be required just for this TF, thus we will concentrate on IPv4 (& dual stack).

Full list of volunteers for dry-run, updated during the meeting and with some email exchanges afterwards:

KIT, CNAF, Imperial College, Glasgow, Bristol. 

Discussion on whether the volunteers are all actually being tested.

KIT - Atlas, one of two possible endpoints is there 
    - LHCb both are there (but read only)

CNAF - Atlas – it's there
     - LHCb – change needed (details communicated)

IC, Glasgow – OK

Bristol – Cédric to check status.

Review of latest probe, framework and results

Discussion on the presentation of results postponed as the appearance will change with the move to pre-prod and SAM3 will become available.

Read-only sites: neither Atlas nor LHCb wish a read-only site to be marked as failed. We thus need the “standard file”, discussed in the previous meeting, to be placed at the sites. Atlas and LHCb agreed to do this.

Discussion with KIT who explained that HTTP PUT for LHCb fails because it does not address a space token. For this to work with their dCache, LHCb must configure some kind of space token as the default (to be used when no ST is specified). Thus, fixing some of the read-only cases will need action from the VOs.

Information for sites

No comments here. Agreed to update it as we go with experience.

Remaining steps before contacting sites

None identified, we can move to the trial runs.

Test run with volunteer sites

List of sites identified above.

Agreed to use GGUS with a small group doing the initial support – Oliver, Georgios, Fabrizo, Cédric, Christophe.

Oliver to circulate a draft process which we will follow and adapt on the basis of experience.

Next meeting

Not scheduled – hopefully before the end of the year, but tbc according to how the trial run goes.