BiLD-Dev
Bi-Weekly "Loyal" DIRAC developers meeting. And, following, the LHCbDIRAC developers meeting.
Join Zoom Meeting
https://cern.zoom.us/j/91083694183?pwd=ZkdDY1I1YkJVc2o3UTdBY1BRZE15UT09
Meeting ID: 910 8369 4183
Passcode: 12345678
One tap mobile
+41432107108,,91083694183# Switzerland
+41315280988,,91083694183# Switzerland
Dial by your location
+41 43 210 71 08 Switzerland
+41 31 528 09 88 Switzerland
+41 43 210 70 42 Switzerland
+33 1 8699 5831 France
+33 1 7037 2246 France
+33 1 7037 9729 France
Meeting ID: 910 8369 4183
Find your local number: https://cern.zoom.us/u/artUfAdNB
Join by SIP
91083694183@188.184.89.188
91083694183@188.185.118.153
Join by H.323
188.184.89.188
188.185.118.153
Meeting ID: 910 8369 4183
Passcode: 12345678
BiLD (Bi-weekly DIRAC Development meeting) – 16/09/2021
At CERN: Federico, Alexandre, Christopher, Christophe, André, Vladimir (in our offices…)
On Zoom: Andrei, Andrii, Cedric, Daniela, Hideki, Simon, Yuji, Janusz
Apologies:
Follow-up from previous meetings
- hackathon 8th September, 7.4.0a1:
- quick py3 hackathon on lbcertifdirac70 box
- no big issues encountered
DIRAC communities roundtable
GridPP:
Daniela
- Running fine, no upgrades
- 3 issues:
- proxy expiring warnings (will make an issue)
- PR 5392 : permission control for accounting (prob PR is OK)
- FC kept hanging (at about 20 requests/seconds) --> SSL ERRORs
- Christopher this is more or less the same threshold that we have seen in LHCb, suggestion is to try to apply #5394
- another suggestion is to simply run more services
ILC/CLIC/FCC:
André
- NTR
LHCb:
Federico+Christophe+Christopher+Alexandre
- Running python3 pilots everywhere (only waiting for VAC sites to update)
- Federico Q: what about VMDIRAC?
- from 7.3 the startup scripts should be changed (do not bother with VMDIRAC used by v7r2).
- Federico Q: what about VMDIRAC?
- Fighting the whole week against overloading of master CS. Initially we thought that moving to py3 pilots was the issue, but we could not relate it. We ended up:
- setting
/DIRAC/Configuration/RefreshTime = 600(the default is 300). - hotfixing https://github.com/DIRACGrid/DIRAC/pull/5394
- ran some “Replica CS” with v7r3 + py3, working fine (and faster)
- setting
- Ran a first LHCbDIRAC hackathon using v7r3p1 as base (py2 server, py3 client and pilots), no issues encountered.
- will soon move this installation to py3 server
EGI:
Andrei
- Moved most of the services to v7r2, one service has to be left to v7r1 (for robot proxy upload).
- a fix is ready, will be made soon.
- Setting up v7r3p1 server with the goal to run http services
- Need to set up WMS and ProxyManager with http for the OpenMOLE community (scala language) – this needs to be developed.
Belle2:
Hideki, Cedric, Ueda
- from previous meeting Try to use BySize plugin - only input by integer GB which is too much. Possible solution is to use floats for the size
- Observed some finished jobs that are again “Running” - asked to post on forum.
BES
Xiaomei
- from previous meeting Asking for the transfers monitoring, this does not exist yet
- in the pipeline but not for now - requirements and a plan of work would be needed
- Chris FTS has a grafana dashboard, if you are using it
- globalDefaults issue from forum: will be fixed
JINR
Igor
- NTR
DIRAC releases
-
v7r2p25:
- purely fixes
-
7.3.1:
- Created the final 7.3 (patch 7.3.1): nothing new wrt the previous pre version, mostly fixes for tornado based services
- NEW: (#5000) Support for Python 3 server installations
- NEW: VMDIRAC separate project is merged into the core DIRAC project
- NEW: (#5042) RSS: Multi-VO mode of operation support
- NEW: (#4903) Added WMS (Jobs) State Machine
- NEW: (#5214) Added HTTPs JobManager, JobMonitor, JobStateUpdate services
- NEW: (#5314) Add a PilotStatus module to clearly define the pilot status in the code
- CHANGE: (#4715) removing env variable DIRAC_USE_NEWTHREADPOOL
- CHANGE: (#4937) removed dirac-agent, dirac-service, dirac-executor scripts (use ‘_’ counterparts instead)
- CHANGE: (#5337) removed src/DIRAC/Core/scripts/dirac-install.py
- CHANGE: (#5110) Dropped Glue2Only and Glue2URLS options from Bdii2CSAgent, Glue2 is now the only way
- CHANGE: (#4303) Removed completely the SystemLogging
- CHANGE: StatesMonitoring agent removed
- CHANGE: (#5313) Docs: replace the theme by the sphinx_rtd_theme
-
7.4.0a1
- Dedicated to Tokens and integration on new IdPs
- Python3 servers installations “defaults”
-
Q: should we re-brand 7.4 to 8.0 (and just skip 7.4)?
- Andrei if we are fully ready with dropping python2, then why not
- after brief discussion: we agree on moving directly to version 8.0, skipping 7.4, and dropping with it python2 support: the consensus is that maintaining python2 and DIRACOS would be more painful
- all 7.4 issues will be moved to 8.0
DIRAC projects
DIRAC:
Issues by milestone:
- v7r1:
- We don’t want to put more effort in this version, so urged to finish off with the existing issues (mostly documentation)
- v7r2:
- NTR
- v7r3:
- NTR
- v7r4:
- will be merged in 8.0
- v8.0:
- NTR
Other issues:
*
PRs discussed:
-
[v7r2] Add pr-sweep action for PR cherry-picking
- Federico if we drop this idea of using
blackformatter, would there be any resistance to adopt this PR/system (from v7r3 or integration)?
* black v7r2, 3, integration- proposal, after discussion:
- adopt black. From release 7.2
- add a pre-commit hook
- test on DIRACGrid-Test, final decision in 2 weeks. Chris Burr volunteered to take care of everything needed.
- proposal, after discussion:
- Federico if we drop this idea of using
-
[v7r2] Helper functions to make working with errors easier
- Federico Considering we’ll never reach a full consensus here, how to proceed?
- Dicussion conclusion: implement naming as per https://github.com/DIRACGrid/DIRAC/pull/5145#issuecomment-844237980
-
from previous meetings
- RFC prepared: https://github.com/DIRACGrid/DIRAC/wiki/Generalization-of-VO-and-Setup-specific-configuration-settings
- Discussed in https://github.com/DIRACGrid/DIRAC/issues/5287
- New draft PR created
- André if the changes are only useful for WebApp, then it’s not worthy, let’s not do changes that are not needed
- The question was raised whether we should keep the notion of the Setup/System Instance in DIRAC. This feature can be considered obsoleted now where separate DIRAC installations can be easily made instead of having separate setups. IF we decide to drop Setups, the configuration will be much simplified and also the implementation of this PR. A new RFC/Discussion will be opened for this issue (Andrei)
- RFC prepared: https://github.com/DIRACGrid/DIRAC/wiki/Generalization-of-VO-and-Setup-specific-configuration-settings
WebApp:
- from previous meetings Problem with empty ComponentMonitoring data - to be checked
- need to use the DIRAC service, not go directly to ElasticSearch
- Started to make PRs to use BaseRequesthandler, for WebApp handlers
- Christophe more documentation is needed
- Next release: 5.0 (python3 only)
Pilot:
- globalDefaults discussion
- Also, connected there was already an old issue Instrument the pilot to pass a GlobalDefaults.cfg file when installing DIRAC Federico will take care
DIRACOS:
- from previous meeting at some point we’ll use the forked elastcsearch open clients, when they will be available (should come at https://github.com/orgs/opensearch-project/repositories)
- frozing to v7r3
DIRACOS2:
- from previous meeting
fts-reston py3?- fts3-rest is added
- the “official” client will be worked “soon”
- Still missing
rucio- 2 issues on their github still open
Documentation:
- https://github.com/DIRACGrid/DIRAC/pull/5304
- generated 1 page for all the commands: https://dirac--5304.org.readthedocs.build/en/5304/CodeDocumentation/Commands/index.html (it was 1 page/command before)
- André I don’t know if I like it. Federico I do like it.
- Christophe does it work also with extensions? I will give you a link for testing
OAuth2:
- from previous meeting Any news from issue WLCG and Token transition – reminders and requirements ?
- meeting tomorrow. It would be good to show a DIRAC version that works with tokens.
- [for next time] Proposal to “move Registry to IdPs”
- e.g. DIRAC Groups == VOMS roles
tornado/HTTPs
- NTR
management
- a few minor PRs to be merged
diraccfg
- NTR
COMDIRAC
- NTR, but Pierre Gay should be ping-ed
DB12
Alexandre
- Port to python3:
- In python3:
long==intand this generates slightly different scores. - We ran DB12 on many different processors: dividing the score by 1.16 seems to correct the issue for Intel processors but not for AMD ones.
- To be decided exactly how to move forward (by next BiLD)
- In python3:
other externals, including Rucio
- Change in the rucio file catalog client itself, requires a config file – this discussion is not specific for DIRAC, though.
- PR: if Cedric approves it, we merge it
Release planning, tests and certification
-
Certification machines
- lbcertifdirac70 machine:
- no changes
- lbcertifdiracoauth machine:
- no changes
- lbcertifdirac70 machine:
-
Next hackathon(s)
- on lbcertifdirac70
AOB and topics from Google forum
- from previous meeting GitHub is introducing “Discussions”: https://github.com/DIRACGrid/DIRAC/discussions/landing. Maybe replace the google forum?
- enabled – bu the google forum is still active.
Next hackathon: on September 23rd:
Next BiLD on September 30th: https://indico.cern.ch/event/1077642/
LHCbDIRAC
- case inconsistency between fs/dfc/bkk: https://mattermost.web.cern.ch/lhcb/pl/fz4f19zum3rj3refkbudnqciqa