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) – 03/03/2022
On Zoom: Federico, Alexandre, André, Andrei, Cedric, Christophe, Christopher, Daniela, Janusz, Ruben, Simon, Hideki, Xiaomei, Ueda
Apologies: Andrii
Follow-up from previous meetings
- hackathon of February 24, 8.0.0a16:
- Tried on lbcertifdiracoauth machine, with board https://trello.com/b/ivqWHmfx/oauthz
- machine was set up according to some steps in https://codimd.web.cern.ch/EKoNwWvZREO4v3s-Brmr8Q?view (not fully finished)
- did not really do much, as could not deploy the release (problem maybe not related to dirac-login/IdP etc). It was anyway an occasion for testing this machine and for seeing in action the issues (see below)
DIRAC communities roundtable
GridPP:
Daniela+Simon+Janusz
- NTR
- ProxyDB issues: upgraded to 7.2.40, will upgrade to 7.3 and then we’ll check again.
LHCb:
Federico+Christophe+Christopher+Alexandre+Ruben
- Moved one production ES to OpenSearch. No issues.
EGI:
Andrei
- from previous meeting Python3 client installation
dirac-configureis very confusing, globalDefaults logic should be added (in the works)- the suggestion is to make a
globalDefaults.jsonfor differentiation (py3 vs py2) and ease of use
- the suggestion is to make a
- from previous meeting Python3 pilots can not be configured to install (COMDIRAC) extensions: this feature should be added to Pilot code.
- Had to switch to Python2 pilots for some communities
- Config of sites WRT to VOs. BDII not always up-to-date, need to check with others, e.g. GOCDB. In contact with “Vapor” portal https://operations-portal.in2p3.fr/vapor/ (that is supposed to aggregate various sources of information)
ILC/CLIC/FCC:
André
- NTR
Belle2
Hideki+Ueda
- Trying to move to v7r2
- Created the BelleDIRAC package for py3, trying to create the extension of the web
- Christopher Look in https://github.com/DIRACGrid/dirac-webapp-packaging
- LHCbWebDIRAC example: https://gitlab.cern.ch/lhcb-dirac/LHCbWebDIRAC
Juno
Xiaomei
- Opened https://github.com/DIRACGrid/DIRAC/discussions/5909
- Wants to use metadata functionalities of DFC.
- Andrei Should be working. Provenance information could added there too.
- we can get together with data manager of Juno, to see what exactly are the use cases. Probably what is already there is good enough, but let’s take this offline.
- Andrei Should be working. Provenance information could added there too.
- https://dirac.readthedocs.io/en/latest/AdministratorGuide/Tutorials/dmsWithTs.html
- suggestion: use for example the
BySizeplugin
- suggestion: use for example the
- Check also https://github.com/DIRACGrid/DIRAC/wiki/Transformation-System-Tutorial
Topics from GitHub/Discussions or Google forum
- Systematic errors in upload operations from a specific site
- should we do some unsetting on our side? prob not.
- Specify a list of CS url in the extension_metadata
- answered
- Strange tag behaviour
- original question have been answered, but there was a follow-up in About SI00, CPUTime, maxCPUTime and what more
- Alexandre can maybe answer
- Andrei for the world outside HEP, these units make no sense. This at some point should be made more user-friendly
- original question have been answered, but there was a follow-up in About SI00, CPUTime, maxCPUTime and what more
- Job priority setting for production users and individual users
- Federico I will try to answer later
DIRAC releases
- v7r2p41:
- minor fix for PublisherHandler
- one commit added to rel-v7r2 for Matcher fix (tags)
- v7.3.18:
- Core
- CHANGE: (#5887) Use importlib.metadata rather than RCSID for getting AgentModule version
- CHANGE: (#5869) Show callstack in service logs when returning S_ERROR
- WorkloadManagementSystem
- NEW: (#5884) Added Account option for SLURM
- CHANGE: (#5850) HeartBeatTime is set every time a status is changed by the job itself
-
- docs changes
- Core
- 8.0.0a15 + 8.0.0a16:
- CHANGE: (#5870) use
DIRAC.Core.Base.Script.Scriptinstead ofDIRAC.Core.Utilities.DIRACScript.DIRACScript dirac-loginfixes- Monitoring
- NEW: (#5807) Added DataOperation to Monitoring
- CHANGE: (#5870) use
DIRAC projects
DIRAC:
Issues by milestone:
- v7r2:
- NTR
- v7r3:
- NTR
- v8.0:
- new dirac-login improvements created based on what seen in the hackthon
- v8.1:
- NTR
Other issues:
- ReplicateAndRegister Operation doesn’t consider SourceSE in FTS3 Mode?
- Simon I’ll give it a go to the proposed solution (by André) in our installation and report later.
PRs discussed:
*
WebApp:
- Several PRs open by Andrii. They should be reviewed (some reviews done)
- Andrii will be absent for a while
- André we can share the code reviews
Pilot:
- from previous meeting Janusz Revived the code for the remote pilot logger to store log messages in a remote SE. Trying to log as early as possible.
- Janusz shared the screen, and showed what he achieved up to now.
- Reused some of the Pilot code that was written in the past
- The logger sends single (JSON) message to a Tornado area. On the server side an agent is in charge of a first post-processing of the messages. Possibilities (to implement):
- log files are closed and uploaded to a Storage Element (being S3 or others… does not fully matter)
- forwards to ElasticSearch the JSON messages as ES documents (or to another backend, but ES seems “natural” now)
DIRACOS2:
- from previous meeting We should be able to specify a DIRACOS2 version: in SystemAdministrator and Pilot we should specify the version
- Andrei DIRACOS2 installers are all in CVMFS now
- Christopher will try that, and code now that everything is there
- Restored Micromamba, not in a release yet, will do it on Monday
Documentation:
- from previous meeting André: diracgrid.org forwards to http://mardirac.in2p3.fr with the docs available
- Builds of the docs in RTD are working again
OAuth2:
- from previous meeting Any news from issue WLCG and Token transition – reminders and requirements ?
- There will be one this afternoon, probs focus on tokens for Rucio
- During hackathon, we revived the idea of externalizing groups management to IdP. This was already discussed in BiLD of 11/11/21 (see also slides from 12)
- has to be exactly 1-to-1 group-IdP
- the CS/registry would need
- Christopher we should probably just make it that the WebPortal behaves like the CLI
- Andrei we should then remove the possibility to change easily the DIRAC group from the WebApp drop-down menu
- basically it’s a logout-login if you want to change the group/scope
- has to be exactly 1-to-1 group-IdP
tornado/HTTPs
- from previous meeting Federico with py3/tornado, could we potentially get rid of of own tornado fork?
- it depends on Python 3.10, but of course we never tried it.
- python 3.10.1 has been released
- but not a trivial thing (first need to try)
- from previous meeting Andrii We can eventually special classes in Configuration and BundleDelivery to pass binary data - to be discussed.
- Federico Trying to have ProxyManager on HTTPs, PR should be checked
management
- Andrei pushed DIRACOS2 installers all in CVMFS
diraccfg
- NTR
COMDIRAC
- NTR
DB12
Alexandre
- Pushed a new version to fix a minor issue with the correction of the norm
- Asked to check for the case of MP pilots
- I will check again the norm
other externals, including Rucio
- from previous meeting Janusz https://github.com/DIRACGrid/DIRAC/pull/5302 has been converted to “real” (it was a draft). Linked to a rucio PR, and waiting for a release. Problem was in the rucio configuration.
- changes in rucio have been merged 2 weeks ago, need to check if they actually made to a release.
- maybe restart the tests
- from previous meeting Andrii there’s no possibility to pass tokens (or proxies) as arguments.
- Cedric I will mention
- no news but there will be a discussion about tokens this afternoon.
- Cedric I will mention
Release planning, tests and certification
-
8.0 release
- the release will be done somewhat soon-ish keeping the AuthZ and token developments that have been done as a technology preview. We might do 1 more hackathon on lbcertidiracoath machine but the release will be effectively certified on lbcertifdirac70 using certificates, proxies, and VOMS.
- we agreed on incorporating the code that has already been written for WebAppDIRAC, and then verify that there is no regression
-
Certification machines
- lbcertifdirac70 machine:
- NTR
- lbcertifdiracoauth machine:
- machine has been updated (mostly puppet, but also its configuration) and it is now usable.
- All HTTPs services added
- Added the configuration for authenticating via external IdPs.
- There’s no connected ES (no need)
- machine has been updated (mostly puppet, but also its configuration) and it is now usable.
- lbcertifdirac70 machine:
-
Next hackathon(s)
- Suggestion is that for the next hackathon to use again lbcertifdirac70
AOB
Next hackathon on March 10th
Next BiLD on March 17h
LHCbDIRAC
- v10r3: deploy board in https://trello.com/b/kzUKdMts/deploy-v10r3
- Moved last ES6 to OpenSearch (https://trello.com/c/ZWZexbH9)
- lbwms: issues on the server side maybe solved
- MONIT: added LHCb dashboards from https://monit.web.cern.ch/ (work in progress)
- added a homepage of organization
- added the WMSHistory dashboard
- few suggestions
- from previous meeting Christopher the BookkeepingDB assumes a certain detector (Run2). This is wrong because it is not anymore the same detector, but is it useful at all?
- Christophe I made a MR changing the way the condition path is constructed. While doing this I found out that there is quite some data that we can’t access
- Christopher it might be that some changes that I have started doing locally are fixing the described issue
- Christophe we should really make the description unique
- Christophe I made a MR changing the way the condition path is constructed. While doing this I found out that there is quite some data that we can’t access
- from previous meeting Alexandre HPCs status: on BSC we finally have a LHCb box
- Update Alexandre Prepared a certification instance (
lhcbvoboccertif03) to push jobs in Mare Nostrum: currently resolving issues and monitoring the machine when we push a large number of jobs. Will prepare a PR once everything will be fine.- 2 PRs (one in DIRAC, one in LHCb) merged, but no release yet
- new beefy machine on the way (for production)
- Update Alexandre Prepared a certification instance (