BiLD-Dev
Bi-Weekly "Loyal" DIRAC developers meeting. And, following, the LHCbDIRAC developers meeting.
Join Zoom Meeting
https://cern.zoom.us/j/91857549918?pwd=S2c1ZmFpYVNDOGk4YjMrZmhOeGROUT09
Meeting ID: 918 5754 9918
One tap mobile
+41315280988,,91857549918# Switzerland
+41432107042,,91857549918# Switzerland
Dial by your location
+41 31 528 09 88 Switzerland
+41 43 210 70 42 Switzerland
+41 43 210 71 08 Switzerland
+33 1 7037 2246 France
+33 1 7037 9729 France
+33 1 8699 5831 France
Meeting ID: 918 5754 9918
Find your local number: https://cern.zoom.us/u/aUn2ex94J
Join by SIP
91857549918@188.184.110.70
91857549918@188.184.89.188
Join by H.323
188.184.110.70
188.184.89.188
Meeting ID: 918 5754 9918
BiLD (Bi-weekly DIRAC Development meeting) – 05/11/2020
At CERN: Again, nobody :/
On Zoom: Federico, Alexandre, André, Andrei, Andrii, Cedric, Christophe, Christopher, Daniela, Hideki, Januz, Marko, Ruslan, Simon, Yuji, Ueda
Apologies:
Follow-up from previous meeting
- Quite some work on DIRACOS2
- PR to conda-forge about DIRAC in Jupyter notebooks? https://github.com/conda-forge/dirac-grid-feedstock/pull/32
- closed
- DIRAC-RucioCatalog plugin coming up (we’ll hear about it today)
- Ran first hackathon last week on v7r2-pre18: https://trello.com/b/vYk0M4KL/v7r2-pre18
- issues:
- Few more JSON serialization fixes, PR open by Christophe
- RSS issues (for v7r1, in fact)
- Still some SingularityCE issues (v7r1)
- …
- goodies:
- python 2 and python 3 clients look fine
- no major issues found
- multi-VO setup looks OK-ish
- new functionalities not yet tried out
- issues:
reminder Marko prepared a container to be used for CVMFS (auto)deploy: https://github.com/DIRACGrid/management/packages/342716 (also in https://hub.docker.com/repository/docker/diracgrid/dirac-cvmfs). For using it from GitHub Actions template in: https://github.com/DIRACGrid/management/pull/19 Andrei + Federico
reminder DIRACOS should be expanded on CVMFS Andrei
DIRAC communities roundtable
GridPP:
- Hopefully by the next meeting should run v7r1
CLIC:
- on v6r22 + DIRACOS
- NTR
LHCb:
- Still on v7r0. Will move to v7r1 ASAP (basically only waiting for RSS fix)
- Fully using M2Crypto (also on master CS and Web boxes).
EGI:
- EGI-ace project approved (continuation of EOS-HUB)
- Using the Conda environment in EGI Jupyter notebooks
- terminal works quite well
- users don’t have certificate inside the terminal: Andrii created command to transform token in proxy (if proxy is uploaded)
- scripts don’t quite work
-
Christopherpackaging is not yet fully finished open issues at if appropriate https://github.com/conda-forge/dirac-grid-feedstock/issues
-
Andreiit might be DIRAC the issue
-
Andreiwould be nice to have COMDIRAC too in conda. It’s supported even if there’s not much to do, developer is there.
-
ChristopherI opened an issue for COMDIRAC for that. Should be pip-installable: https://github.com/DIRACGrid/COMDIRAC/issues/72
- terminal works quite well
- 10 days ago Andrei gave DIRAC EGI Webinar: quite well attended
France Grilles:
- Will be merged in EGI installation at some point
- will take time
Belle2:
- Started discussing about migration to v7r0. It will not come in this year. Now the focus is on Rucio: migration will be done on v6r22+DIRACOS
JINR:
- NTR
Juno, BES3:
- NTR
Current situation
DIRAC
- v7r0p37:
- Minor fixes
- v7r1p16:
- Minor fixes
- v7r2-pre18:
- Many fixes related to python 3, including switch for installing python 3 client
WebApp:
-
fixed the versioning:
rel-v4r0branch for DIRAC versions:- v6r22
- v7r0
- v7r1
rel-v4r2branch for DIRAC version:- v7r2
integrationfor future (v7r3)
-
rel-v4r1branch has been retired -
Use ngnix for data transfer? in progress
Pilot3:
- NTR
DIRACOS:
- xroot5? no news
- ldap3 should be added (PR open)
-
Hidekiis
lessworking? > from v1r14 - next version can include
tornado
DIRACOS2:
-
Added repo to DIRACGrid https://github.com/DIRACGrid/DIRACOS2
- now functional
- last hackathon still used Chris’ DIRACOS repo, need to change
dirac-installfor update the link: will be done this week
-
https://anaconda.org/diracgrid/ is for “specific” packages
-
FedericoWhy M2Crypto is here?
-
used by the unit tests (for v7r1 CE, for v7r2 no need?)
-
-
-
Other “standard” packages not distributed via (ana)conda are added to conda-forge.
VMDIRAC:
- NTR
Documentation:
- NTR
OAuth2:
- WIP
tornado/HTTPs
- NTR
management
diraccfg
- NTR
other externals
- NTR
Release planning, tests and certification
-
Certification machine updated to be multi-VO (task: https://github.com/DIRACGrid/DIRAC/issues/4631)
- André GridPP VOMS info should be there now (added in central puppet)
- tests not updated yet to be multi-VO
- Federico Will probably update ES to ES7, but first need to sort out a couple things
-
Next hackathon November 12th
- Already called in https://indico.cern.ch/event/970877/
-
v7r2:
- several PRs open, mostly for
- “minor” fixes
- removing depracated code
- e.g. MultiProcessorSiteDirector removed
- ProxyDB backward compatibility
- tests updates
- Removed MySQL 5.6 tests, set MySQL 8 as default
- several PRs open, mostly for
Weekly development(s) focus
topic: DIRAC-Rucio integration
GridPP (pres on the agenda) [Janusz]
- took long time, as was waiting for multi-VO Rucio catalog implementation
- RucioFileCatalog plugin activated for v7r1 (in test)
- nicely integrate with DIRAC tools. Minimal modifications.
- Storage Element definitions (DIRAC-Rucio) not fully sorted out
- to API or not to API?
- has to include rucio in DIRACOS
Christopherit’s a no-brainer packaging, as rucio is pure python
Christopheit’s properly packaged, even DIRACOS 1 can package that easily
Cedricpython2.7 and 3 supported, 2 might be dropped in the future months - to REST or not to REST?
- licensing issues?
- if we use the python bindings there’s no licensing issues.
General consensus: using python bindings instead of REST look like a better option
Belle2 (pres on the agenda) [Cedric]
- main issue was mapping the Rucio scopes to the namespace
- using python bindings with DIRACOS
- need to install the rucio.cfg (1 per VO) - this can’t be skipped, won’t go through DIRAC
- 1-to-1 mapping between SE and RSE (via agent)
JanuszConfiguration: can’t I eliminate it completely? It can all go in DIRAC. It’s possible by a configuration parameter in Rucio client
Andreialso pilots are clients so we would need to ship this file too. We should NOT have it needed
Cedricit’s still needed ATM. Changes to the API are needed. Alternatively it can be created on the fly.
General questions/comments:
IgorCan different VO under DIRAC use different single VO Rucios? => should be, but never tried, nor sure
IgorDoes DIRAC accounting work correctly with Rucio plugin?
Andrei+Uedashould be yes, it’s just another catalog, for synchronous operations (if you use Rucio only as file catalog)
Janusz+Uedaif you use Rucio as data manager (for moving datasets) then no.
ChristopheRucio has their own rules, there’s no way DIRAC knows about them
Christophewith DIRAC usinghttpsthe interactions between one and the other will be much simpler. But this is not for tomorrow.
FedericoMore discussions via mail. Only 1 implementation.
DIRAC: current PRs and tasks being worked on, or topics from Google forum
PRs
- BDII and SE
- agreed to stop looking for SEs inside BDII Federico
Issues
- v7r0:
- Nothing in there
- v7r1:
- Several tasks in there still, so this is to be considered an actively developed release.
- URGENT: MultiVO-RSS breaks interface
-
JanuszI will look into it => put the
voat the end for backward compatibility for a quick fix.
- v7r2:
- mostly non-urgent
- v7r3:
- Not discussed
Other issues:
- HTCondor TimeLeft module
- the information from CERN only are better than nothing Alexandre will create the PR.
- runit in dirac is old
- topic will be closed
Topics from Google forum
- NTR
AOB
Next hackathon in 1 week.
Next BiLD in 2 weeks.
LHCbDIRAC
- Waiting for RSS fix on v7r1 before releasing v10r1.
- Christophe I will circulate a document for changes for the LHCbDIRAC certification machine.