generated from sig_core/wiki-template
6.6 KiB
6.6 KiB
SIG/HPC meeting 2023-12-14
Attendees:
* Sherif Nagy
* Neil Hanlon
* Matt Bidwell
* Rich Adams
* Chris Simmons
* Jeremy Siadal
Follow ups
- No movement on wiki yet, maybe over break
- Cnode Kernel - no movement
- Compare spec files for Warewulf vs OpenHPC - Done!
- Thank you Sherif
- Building warewulf4 for rocky 8 and rocky 9
- Can we keep this name w/ openhpc?
- Chris - next release will also rename warewulf to warewulf3 to distinguish
- Request SB Certs for HPC cnode kernel from Security - Requested
- Requested - No update
- Investigate what resources are required for testing - Neil
- Not done yet
- Create POC script to create tickets when new slurm is available - Neil
- No movement
- Sherif waiting to hear from Jeremy about Intel GPU drivers
- Should have heard from them. Jeremy will follow up with them to see what happened
Discussions
- last meeting of 2023; skip Dec 28th. next meeting will be Jan 11 -- needs announcing
- Happy holidays!
- slurm naming - slurm22 / slurm23 / slurm24
- slurm24 is coming out soon
- plan to support whatever schedmd is supporting -- two most recent releases
- Testing resources for SIG/HPC
- NVidia V100s - OK?
- Cannot test MIG (multi-instance GPU) with that device
- Sherif can take some of these after they decomm their current HPC, but not sure on timeframe
- Need a place to host these, maybe RESF can do something
- Neil is tracking this, to have better update in January
- NVidia V100s - OK?
- chris is working on testing for different schedulers
- Warewulf -- OpenHPC needs to make naming more consistent
- will remove warewulf4 from builds once it's in Rocky and other openhpc distros
- Rocky not worrying about v3, openhpc will continue providing that
- Slurm / pmix support
- on for rocky 9 branch
- there is a pmix5, but ... it's broken. Chris is looking at this over holiday break
- rocky only has pmix 3.2, so if we need new features we may need to build and release in the SIG
- newer versions (4) are backwards compatible, or, are supposed to be
Action items
- Sherif to finish/complete work on the wiki
- Decide what is being put into cnode kernel, what is being removed - Jeremy
- Request SB Certs for HPC cnode kernel from Security - Requested
- Requested
- Create POC script to create tickets when new slurm is available - Neil
- Change warewulf -> warewulf3 in next openhpc release - Chris
- Announce meeting cancelations for December - Neil/Sherif
- Look into building pmix4 for rocky and building slurm23.11 w/ pmix support - Sherif
- Follow up with Intel Driver team - Jeremy
Old business
2023-11-30
- Sherif to finish/complete work on the wiki
- Not done
- Decide what is being put into cnode kernel, what is being removed - Jeremy
- No updates
- Request SB Certs for HPC cnode kernel from Security - Requested
- Requested
- Compare spec files for Warewulf vs OpenHPC - Sherif
- Not done yet
- Investigate what resources are required for testing - Neil
- Not done yet
- Create POC script to create tickets when new slurm is available - Neil
2023-11-15
- Sherif to finish/complete work on the wiki
- Not done
- Sherif to add Jeremy and Chris to gitusers and sig_hpc - Done
- Decide what is being put into cnode kernel, what is being removed - Jeremy
- No updates
- Request SB Certs for HPC cnode kernel from Security - Requested
- Requested
- Compare spec files for Warewulf vs OpenHPC - Sherif
- Not done yet
- Investigate what resources are required for testing - Neil
- Not done yet
2023-11-02
* Sherif to work on abit on the wiki - Not done
* Sherif to add Jeremy and Chris to the git user groups
2023-10-19:
* Sherif to create kernel repo for kernel HPC, kernel-hpc-node, called now kernel-cnode - Done -
* Jeermy, to get the ball rolling with intel GPU driver
* Stack, Fix the slurm rest daemon and integrated it with openQA
2023-10-05:
* None for this meeting, however we should be working on old business action items
2023-09-21:
* Sherif: Get the SIG for drivers
* Sherif: Check the names of nvidia drivers "open , dkms and closed source"
* Chris: Bench mark nvidia open vs closed source
2023-09-07:
* Sherif: Reaching out to AI SIG to check on hosting nvida that drivers that CIQ would like to contribute - Done and waiting to hear from them -
2023-08-24:
* Sherif: To push the testing repo file to release package
* Sherif: testing / merging the_real_swa scripts
2023-08-10:
* Sherif: Looking into the openQA testing - Pending
2023-07-27:
* Sherif: Reach out to jose-d about pmix - Done, no feedback yet -
* Greg: to reach out to openPBS and cloud charly
* Sherif: To update slurm23 to latest - Done -
2023-07-13:
* Sherif needs to update the wiki - Done
* Sherif to look into MPI stack
* Chris will send Sherif a link with intro
2023-06-29:
* Sherif release slurm23 sources - Done
* Stack and Sherif working on the HPC list
* Sherif email Jeremy, the slurm23 source URL - Done
2023-06-15:
* Sherif to look int openHPC slurm spec file - Pending on Sherif
* We need to get lists of centres and HPC that are moving to Rocky to make a blog post and PR
2023-06-01:
* Get a list of packages from Jeremy to pick up from openHPC - Done
* Greg / Sherif talk in Rocky / RESF about generic SIG for common packages such as chaintools
* Plan the openHPC demo Chris / Sherif - Done
* Finlise the slurm package with naming / configuration - Done
2023-05-18:
* Get a demo / technical talk after 4 weeks "Sherif can arrange that with Chris" - Done
* Getting a list of packages that openHPC would like to move to distros "Jeremy will be point of contact if we need those in couple of weeks" - Done
2023-05-04
* Start building slurm - On going, a bit slowing down with R9.2 and R8.8 releases, however packages are built, some minor configurations needs to be fixed -
* Start building apptainer - on hold -
* Start building singulartiry - on hold -
* Start building warewulf - on hold -
* Sherif: check about forums - done, we can have our own section if we want, can be discussed over the chat -
2023-04-20
* Reach out to other communities “Greg” - on going -
* Reaching out for different sites that uses Rocky for HPC “Stack will ping few of them and others as well -Group effort-”
* Reaching out to hardware vendors - nothing done yet -
* Statistic / public registry for sites / HPC to add themselves if they want - nothing done yet -