WLCG DOMA BDT Meeting
Topic: WLCG DOMA BDT Meeting (twiki)
-
-
4:30 PM
→
4:35 PM
News 5m
-
4:35 PM
→
4:45 PM
Tape REST access 10mSpeaker: Mihai PATRASCOIU (CERN)
-
4:45 PM
→
5:00 PM
Transfers with tokens 15mSpeakers: Petr Vokac (Czech Technical University in Prague (CZ)), Francesco Giacomini (INFN CNAF)
CERN FTS pilot & tokens - progress
- Now configured also to support ATLAS token issuer
- successful transfers test between dCache (PRAGUELCG2) <-> EOS (CERN) <-> StoRM (INFN-T1)
- audience was not configured correctly on EOS ATLAS (already discussed with EOS Ops)
- different StoRM behavior for "stat" operation
- StoRM:
storage.modify
doesn't allow stat operations while this is allowed in dCache & XRootD- preferred solution
storage.create
&storage.modify
should allow "stat" operation- StoRM code needs to be updated
- currently FTS needs
storage.read
+storage.modify
when StoRM is HTTP-TPC destination- releasing fixed StoRM & deploying at sites takes time and DC24 is already quite close
- Rucio will be able to use workaround with read+modify for DC24 transfers
- consistency of WLCG JWT profile implementation is certainly the important thing
- posix filesystems also would typically not require r for w
- requiring read for "stat" is a bit confusing
- preferred solution
Operational model for tokens during DC24
- CERN would like to know rates, who will be included, what are our plans with token related mini-challenges
- asked for a document that can be shared between iterested parties
- would like to understand both ATLAS and CMS plans
- Rucio developers mentioned situation will be clear within two weeks once we have initial implementation for full chain with tokens
- ATLAS and CMS behavior should be same by design (not much flexibility in Rucio token configuration in initial release that will be used in DC24)
- Now configured also to support ATLAS token issuer
-
5:00 PM
→
5:10 PM
Packet marking 10mSpeakers: Marian Babik (CERN), Shawn Mc Kee (University of Michigan (US))
Succesfull SC23 demo
- Packet marking @300Gbps (see screenshots attached)
- Tested integration with ESnet high-touch service, dataset to be published later this week
- InMon: SFlow-RT readout of the packet marking from network equipment worked fine (screenshot attached)- New fireflies dashboard showing data from a production storage (at UNL):
. - https://public.stardust.es.net/d/b8dddac0-5b24-4739-9c8d-e88a05c1344f/scientific-network-tags3a-rande-dashboard?orgId=1&from=now-12h&to=now
- There is additional data available that is not yet shown both from fireflies as well as from Esnet, e.g. duration of the flows, transfer rates, etc.Developments
- flowd - developed support for packet marking on multiple NICs (tested succesfully @ SC23)
- plan to create separate packages for flow labeling and packet marking (flowd and flowd-ebpf) - different deps + EPEL requirements
- following SC23 tests we will also refactor the way fireflies are processed and how additional data is added
- xrootd http-tpc support was released
- FTS, Rucio and gfal2 - waiting for deploymentdCache fireflies
- dCache AGLT2 tests with fireflies - discovered problems also in latest 9.2.5
- we hope that this could be fixed in next release (december?)
- enabling fireflies is simple - just flip one boolean value
XRootD & support for flowd activity
- available since XRootD 5.6.3
- already available in EPEL
- work in progress with OSG packages
- this functionality is already included in EOS that is now being deployed at CERN
- Rucio needs to be updated to pass activity for SciTags to FTS-1829
- actually seems to be already implemented
-
5:10 PM
→
5:25 PM
WebDAV Error Message Improvement Project & unified error message format 15m
Discuss with experts improvements in the error messages produced by failed transfers.
https://twiki.cern.ch/twiki/bin/view/LCG/WebdavErrorImprovementSpeaker: Stephan Lammel (Fermi National Accelerator Lab. (US))Gfal upload / download disknode hostname (DMC-1390, rucio#5980), ATLAS would like to have these details in Rucio traces / Monit data. For HTTP-TPC we might get this info from performance markers in the future, but we would like to have this information available also for simple upload/download.
- development that's not directly related to DC24 will happen only after this event.
- it seems it should not be difficult to provide this functionality (but first FTS developers needs to think what would be the best way to provide real souce/destination hosts)
-
5:25 PM
→
5:30 PM
AOB 5m
-
4:30 PM
→
4:35 PM