Present: Brian, Gareth, John H, Lukasz, Matt RB, Raja, Sam, Tom, Winnie, Jens, Matt D, Steve, Elena, Ewan

0. Operational blog posts: Ewan's issue was a non-bugfixed DPM: there are
multiple releases.  This one had a problem with files of a not complete block
size in which case it sent and empty packet which confused the other end.

Issues with fetch-crl on SL5?  Certain CAs had their CRLs "expired" which
causes the certificates from those CAs to be rejected (worse things could
happen too)

GFAL issue on CASTOR (Brian): WLCG baseline should be 2.9.2 but the
twiki currently says 2.7.8.  COuld usefully be tested on the UIs first before
being rolled out to the WNs.

MyProxy and GridFTP: Globus advisory, not a big problem.


"Small" and new VOs.

LIGO: successfully ran DIRAC jobs on SouthGrid.  Does silence mean all is
well, or?  Paul is probably doing something else at the moment, and will
return to the good stuff (aka GridPP) later.  Sam will ask Paul about joining
the support list.  CERN VMs seemed to work.  Imperial however, didn't work for
Tom either so that's maybe why it didn't work for Paul.

DiRAC is progressing.  Brian can do a 'dirty' lcg-cp (as RAL's DNS and BDII
(aka CIP) are not yet set up).  However, DNS has to come first because once
published, ops will start testing the advertised endpoint.  And the DNS entry
is needed as it will be part of the SURL.

Should they use LFC or some other FC?  CERN have stopped using LFC, but there
is no production DFC for example; as DFC can talk to LFC backend, it seems
best to use LFC.


From hepsysmen and women, and other things:

Backup into RAL of data from Sussex revisited?  Order of 100TB (Matt RB).

EUDAT also wants to do HTTP(S)->gw->GridFTP where the gw handles the
certificate.  Similar to US science gateways.  GO may have precisely this.

WebFTS (Brian) more stable; we only have the CERN one and need the RAL one
(RAL had one but it was shut down).  How else to manage endpoints for dCache
and CASTOR?  Use WebDAV?  GO does a (Grid)FTP 'dir' which doesn't work on
CASTOR.

Glasgow had looked at disk read ahead settings on disk servers.
Recommendations for RFIO previously had changed with the change in access
patterns (specifically random/scatter vs reading the whole file).

There is a new WG on 'federated' HTTP??  Looking at deployment.

There is a requirement to have the checksum generated at rest instead of in
flight, even if it means another read process?  So looking at async checksum
and also monitoring and reporting.


Paige Winslowe Lacesso: (10/06/2015 10:05:51)
please repeat version!
Samuel Cadellin Skipsey: (10:06 AM)
Winnie: dpm-dsi-1.9.3-2.el6.x86_64 is the bugfix version (1.9.3-1 has the bug)
Ewan Mac Mahon: (10:07 AM)
And do we know yet whether it does?
Matt Doidge: (10:07 AM)
I had this forwarded to me from the globus lists yesterday, regarding the recent logjam vulnerbility:
https://support.globus.org/entries/94083138
Paige Winslowe Lacesso: (10:08 AM)
Thanks, we're running 1.9.5-3
Ewan Mac Mahon: (10:08 AM)
Because if anyone wants me to upgrade al my SL5 servers in a massive hurry, they're going to be disappointed.
Raja Nandakumar: (10:08 AM)
https://ggus.eu/?mode=ticket_info&ticket_id=113766
LHCb saw the problem in SARA
Maybe a good idea for all sites to update all the WNs to the latest fetch-crl version ...
Ewan Mac Mahon: (10:22 AM)
The being on the list isn't really optional, but we can still be polite about telling him that he absolutely has to join it or else.
:-)
Argh!
Gridmap file surely?
Which you write with a text editor.
Because it only needs like one user.
Jens Jensen: (10:32 AM)
"normally upset" :-)
Matt Raso-Barnett: (10:33 AM)
hi, sorry my mic is a problem still
yes this is correct
i haven't followed up yet though
but yes of order couple hundred TBs
I expect STFC funded, but will talk to them this week
i'm keen to investigate this though
particularly in conjunction with experinmenting with lustre changlogs
so we use changelogs to work out cleverly what needs archiving
Ewan Mac Mahon: (10:34 AM)
We should summarise the situation so far on the storage list
And then talk about it on the storage list,
Yay lists.
Matt Raso-Barnett: (10:35 AM)
yes
Ewan Mac Mahon: (10:35 AM)
\o/
Matt Raso-Barnett: (10:36 AM)
if I talk to them and then I can write to the list with roughly what we are looking to do, and perhaps then work out a plan from there
Ewan Mac Mahon: (10:45 AM)
Also, USB headsets are great, btw.
Samuel Cadellin Skipsey: (10:46 AM)
https://twiki.cern.ch/twiki/bin/view/LCG/HTTPStorageRequirements
Jens Jensen: (10:52 AM)
:-)
Ewan Mac Mahon: (10:53 AM)
Aren't we basically already a distributed testbed with a bunch of different solutions?
Tom Whyntie: (10:56 AM)
Thanks, bye