✨ NEW: Add AiiDA v2.X support for eiger
#80
Conversation
mbercx
force-pushed
the
new/aiida-v2-support
branch
from
5207871
to
50d8f5e
Compare
There was a problem hiding this comment.
Hi @mbercx, and thanks for the PR 👍 .
I have a few suggestions, and comments:
- In the current setup, we still accept codes and computers in the "old" AiiDA-1.x style. They get automatically converted into AiiDA-2.x style by the script. See the output here.
- I see you've put codes setup into
codes-installed. At the moment, this won't work as the parser is not reading that folder.
Overall, I would be happy if we could convert the "source" of codes/computers to AiiDA-2.x. But then we need to update the migration script to do 2.x -> 1.x
| transport: core.ssh | ||
| scheduler: core.slurm |
There was a problem hiding this comment.
I don't think this is needed right now. We have an automated conversion tool that does that.
| #SBATCH --partition=normal | ||
| #SBATCH --account=mr0 |
There was a problem hiding this comment.
Instead of deleting those lines, I would rather create a new computer named multicore
There was a problem hiding this comment.
I would remove "eiger" from the name. I think, the path already suggests that this is for eiger.
There was a problem hiding this comment.
For sure, thanks! Still had to clean that up, this was just me testing my local files.
We discussed that some time ago and agreed to provide a migration tool instead. Pinging @unkcpz in case he has anything to add here. |
|
Ah, sorry, was still also in a draft state because I was playing around with it. Basically I constantly have to create new projects and wanted to stop copying my YAML files and see if I can get a good flexible setup via the code registry instead.
You mean for AiiDAlab, right? But don't we also just want "regular" users to use the registry? |
The migration script is provided in this repository to give two versions of JSON database for AiiDAlab use cases. @mbercx you are correct about this. I think since we gradually at the last stage of moving everything to v2, for this repo it makes more sense to having yaml file in v2 style and generate v1 format backward.
This becomes a barrier to use even. how many users on Earth are actually using it to set up? This is also a very key point for usability. The current structure didn't provide convenience but more confusion in my sense. For this PR itself, I think @mbercx you can just keep using the v1 format as other setups. In AiiDAlab then it can be founded and "potentially" be used. P.S I have a very clear developing path how to support the template for code/computer setup and also having it used in AiiDAlab (which should be even more flexible and easy to support). It requires at least 0.5 person month of work approximately, let's see when I can find time to start that. |
|
It indeed isn't practical. ^^ I was thinking of integrating the registry with the Of course there I could also use this migration script. We should perhaps discuss how to integrate these efforts. I agree that perhaps we should invert the migration approach though, i.e. provide the YAML files for v2 and allow the script to downgrade them.
Great! Happy to discuss this. I think having an easily accessible and smooth functioning registry will help a lot with making things easier for new users within our HPC ecosystem. |
mbercx
force-pushed
the
new/aiida-v2-support
branch
from
41235dd
to
09422b2
Compare
mbercx
force-pushed
the
new/aiida-v2-support
branch
from
53325fd
to
3214d1f
Compare
for more information, see https://pre-commit.ci
| work_dir: /scratch/e1000/{username}/aiida/ | ||
| shebang: '#!/bin/bash' | ||
| mpirun_command: srun -n {tot_num_mpiprocs} | ||
| mpirun_command: srun -n {tot_num_mpisprocs} |
There was a problem hiding this comment.
| mpirun_command: srun -n {tot_num_mpisprocs} | |
| mpirun_command: srun -n {tot_num_mpiprocs} |
| look_for_keys: true | ||
| timeout: 60 | ||
| allow_agent: true | ||
| proxy_jump: '' |
There was a problem hiding this comment.
It only works for me when adding this.
| proxy_jump: '' | |
| proxy_jump: 'ela.cscs.ch' |
|
Please also add code for |
Just updated the Eiger computer and code setup files for a bunch of QE codes. Tested in the field, except for the configuration. Still a draft PR for this reason, as well as the fact that we should probably have a separate configuration directory for v2.X computers. For codes I think it's a good idea to have different folders for the different types, which I already integrated here.
@yakutovicha what do you think? From an AiiDAlab infrastructure point of view, how would you provide support for v1 & v2?