Choose timezone
Your profile timezone:
BiLD-Dev
→
Europe/Zurich
2-R-014 (CERN)
2-R-014
CERN
Description
Bi-Weekly "Loyal" DIRAC developers meeting.
And, following, the LHCbDIRAC developers meeting.
Bi-weekly DIRAC Development meeting -- 16/01/2020
At CERN: Federico, Zoltan, Christophe, Alexandre, Christopher
On Vidyo: Hideki, Daniela, Simon, Janusz, Philippe, Luisa
Apologies: Andre', Marko
Follow-up from previous meeting
- management repo has been pushed by Federico in DIRACGrid
- creates the docker images used for integration tests, upload them to GitHub package registry and docker-hub too
- creates an image for running dirac-distribution and create DIRAC tarballs
- doc for creating DIRAC releases updated accordingly
DIRAC communities roundtable
- LHCb:
- Using DIRAC doc tools for creating LHCb documentation
- Belle2:
- NTR
- GridPP
- NTR apart from VMDIRAC issues (see forum)
- CTA
- moved to v7r0 before the end of year.
- pilot issues: due to some issues, highlighted in messages on the forum, that originated a few Pull Requests, right now for the production setup the server is using v7r0 and the pilots v6 (with pilot2)
- staging issues. Unclear, maybe not release-dependent?
- moved to v7r0 before the end of year.
Current situation
DIRAC
- v6r22:
- NTR
- v7r0:
- v7r0p10 is the last one
- several fixes put in this one, especially for WMS (faster HtCondorCE and ARC) and pilot (originated from CTA issues)
- v7r0p10 is the last one
- v7r1:
- NTR
- v7r2:
- NTR
VM:
- Andrei should answer to https://groups.google.com/forum/#!topic/diracgrid-forum/nghBih6Y7KA
- or GridPP might make a first PR for trying to fix
WebApp:
- NTR
Pilot3:
- PR https://github.com/DIRACGrid/Pilot/pull/100 targeting master with several things:
- moved from travis to github actions
- support for several python versions: 2.6.9, 2.7.5, 2.7.13, 3.6.8
- discovered that 2.6.9 was NOT supported
- 3.6.8 is supported from this PR on, but pending adaptation of dirac-install to python 3
- added shellcheck
- the test can/should be done together with https://github.com/DIRACGrid/DIRAC/pull/4392
DIRACOS:
- new version v1r5, in which:
- git has been added (needed for master CS since v7r0)
- Updated version of MySQL to latest client of 5.7 (5.7.21 to 5.7.28), in order
- this created unexpected issues in case the sever is MySQL version 5.6. For LHCb this has been solved by disabling the encryption server side
- [communities] some communities apparently running old MySQL versions. It has been reminded that the 2 server versions supported are 5.6 and 5.7. When MySQL 8 will be supported "officially", we will drop the support of 5.6.
- Added singularity
- will allow sandboxing and better isolation
- Some issues discovered yesterday (trying to use the nss from the system), more details in https://github.com/DIRACGrid/DIRACOS/issues/112
Documentation
- NTR
OAuth2:
- NTR
tornado and other externals
- JSON encoding PR is moving, being updated
- OpenSSL version: we are stuck with version 1.0, as 1.1 removed the possibility to specify an environment variable for RFC proxies. A PR has been created but not yet merged.
Release planning, tests and certification
Release planning:
- v6r22 series
- Nothing planned
- v7r0 series
- NTR
- v7r1 series
- NTR, apart from several PRs ready to be merged
- Integration (and v7r2)
- Some PRs ready to be merged, waiting for Andrei
Certification process:
- though a github action workflow on the https://github.com/DIRACGrid/management repo, the images used for the tests are now created once per week and uploaded to github package repository, and to docker hub.
- the PR https://github.com/DIRACGrid/DIRAC/pull/4377 will also create (once merged) an "integration" tarball, and a "master" tarball -- this last one should be added by Federico.
- used the new puppet module for new DIRAC certification machine lbcertifdirac70, which is fully operational
Weekly development(s) focus
- Suggestion for a new "Completing" status:
- Philippe raised a point about "completed" jobs, that sometimes, depending on certain condition (e.g. when being killed within finalization) stay in the "completed" status forever if they don't have any associated requests. The suggestion is to use instead a "completing" status by default, and use the "completed" status only for those jobs that effectively have some requests. If this is the case, then the StalledJobAgent could work on statuses "Running" and "Completing". Issue https://github.com/DIRACGrid/DIRAC/issues/4412 open for dealing with this case. Philippe will try a first implementation, targeting v7r1.
- payload environment isolation
- Luisa brought up the subject after moving to v7r0. There's no way to avoid from a community to verify that their payload is started within a proper environment. DIRAC provides the tools for doing it, but it's responsibility of the users to start their payloads in a sanitized environment (see e.g. https://groups.google.com/forum/#!topic/diracgrid-forum/TfiRouxA0XA)
DIRAC: current PRs and tasks being worked on, or topics from Google forum
- v7r0:
- nothing specifically discussed
- v7r1:
- the PR for removing the SEPrefix should be ready to be merged
- the PR added an env variable for replacing DIRAC ThreadPool with native python ThreadPoolExecutor is also ready
- v7r2:
- PRs are now passing the tests because there's an "integration" release
AOB
Next DUW at KEK, 25-29 May 2020. Registration is now open.
Next BiLD in 2 or 3 weeks, depending if in 2 weeks from now we will have a hackathon or not (we should, probably)
---------------------------------------------------------------------------------------------------------------
LHCb
- https://docs.google.com/document/d/1U4DR-n7Omv4oKOmUaqvXkCTazsgGJG1ybfZYmat8Z9Y/edit?ts=5d8241f4
- Deployment: started the new way to deploy to CVMFS also production releases, in parallel with the old one (which will keep being the default)
- full switch from v10r0
- LHCbDIRAC docker container: PR merged, was failing to upload, another PR created with hopefully a fix.
There are minutes attached to this event.
Show them.