Make MPI work #55

2023-06-16T17:11:54+02:00

rarias commented

2023-06-16 17:11:54 +02:00

(Migrated from pm.bsc.es)

Fixes #16

For now we only add MPICH, as I need to start launching MPI tests for ovni.

Although, we still need to copy the closures to each node manually:

hut$ nix shell jungle#hut.bsc.osumb
hut$ nix copy --to ssh://owl1 $(which osu_bw)
hut$ nix copy --to ssh://owl2 $(which osu_bw)
hut$ srun -N2 osu_bw
# OSU MPI Bandwidth Test v7.1
# Size      Bandwidth (MB/s)
# Datatype: MPI_CHAR.
1                       1.30
2                       2.61
4                       5.23
8                      10.45
16                     20.62
32                     41.07
64                     82.23
128                   158.42
256                   315.83
512                   622.96
1024                 1183.65
2048                 2062.81
4096                 3090.81
8192                 4190.01
16384                4826.58
32768                5583.63
65536                7160.62
131072              10715.51
262144              11915.88
524288              12216.00
1048576             12303.94
2097152             12347.91
4194304             12384.89

We should have a shared nix store in the compute machines.

Fixes #16 For now we only add MPICH, as I need to start launching MPI tests for ovni. Although, we still need to copy the closures to each node manually: ``` hut$ nix shell jungle#hut.bsc.osumb hut$ nix copy --to ssh://owl1 $(which osu_bw) hut$ nix copy --to ssh://owl2 $(which osu_bw) hut$ srun -N2 osu_bw # OSU MPI Bandwidth Test v7.1 # Size Bandwidth (MB/s) # Datatype: MPI_CHAR. 1 1.30 2 2.61 4 5.23 8 10.45 16 20.62 32 41.07 64 82.23 128 158.42 256 315.83 512 622.96 1024 1183.65 2048 2062.81 4096 3090.81 8192 4190.01 16384 4826.58 32768 5583.63 65536 7160.62 131072 10715.51 262144 11915.88 524288 12216.00 1048576 12303.94 2097152 12347.91 4194304 12384.89 ``` We should have a shared nix store in the compute machines.

rarias commented

2023-06-16 18:26:33 +02:00

(Migrated from pm.bsc.es)

added 8 commits

09d59fb9...316ea18e - 2 commits from branch master
684d5e41 - Add coments in slurm config
ab048553 - Add mpich overlay
5b41670f - Replace mpi inside bsc attribute
17245354 - Use explicit order in overlays
6402605b - Add missing parameter to extend
6c10933e - Set mpi to mpich by default in bscpkgs

Compare with previous version

added 8 commits <ul><li>09d59fb9...316ea18e - 2 commits from branch <code>master</code></li><li>684d5e41 - Add coments in slurm config</li><li>ab048553 - Add mpich overlay</li><li>5b41670f - Replace mpi inside bsc attribute</li><li>17245354 - Use explicit order in overlays</li><li>6402605b - Add missing parameter to extend</li><li>6c10933e - Set mpi to mpich by default in bscpkgs</li></ul> [Compare with previous version](/gitlab/rarias/jungle/-/merge_requests/11/diffs?diff_id=8579&start_sha=09d59fb99cdb58205c110aa617bca06fbdc867e0)

rarias commented

2023-06-16 18:35:44 +02:00

(Migrated from pm.bsc.es)

added 1 commit

c76bfa7f - flake.lock: Update

Compare with previous version

added 1 commit <ul><li>c76bfa7f - flake.lock: Update</li></ul> [Compare with previous version](/gitlab/rarias/jungle/-/merge_requests/11/diffs?diff_id=8582&start_sha=6c10933e80bdfdbec6438e583df4871c8af67a14)

rarias commented

2023-06-16 18:44:07 +02:00

(Migrated from pm.bsc.es)

marked this merge request as ready

marked this merge request as **ready**

rarias commented

2023-06-16 18:44:08 +02:00

(Migrated from pm.bsc.es)

changed the description

rarias commented

2023-06-16 18:47:47 +02:00

(Migrated from pm.bsc.es)

changed the description

rarias commented

2023-06-16 18:47:48 +02:00

(Migrated from pm.bsc.es)

requested review from @arocanon

rarias commented

2023-06-20 12:40:08 +02:00

(Migrated from pm.bsc.es)

added 1 commit

60077948 - Add osumb to the system packages

Compare with previous version

added 1 commit <ul><li>60077948 - Add osumb to the system packages</li></ul> [Compare with previous version](/gitlab/rarias/jungle/-/merge_requests/11/diffs?diff_id=8601&start_sha=c76bfa7f86f853f431a4b3d34ade3113394032e9)

rarias commented

2023-06-20 12:51:41 +02:00

(Migrated from pm.bsc.es)

mentioned in merge request !12

arocanon commented

2023-06-23 11:02:16 +02:00

(Migrated from pm.bsc.es)

Is this benchmark that essential as to make it available in all nodes by default?

arocanon commented

2023-06-23 11:02:18 +02:00

(Migrated from pm.bsc.es)

Looks good to me! Thanks for keeping me up to date!

arocanon commented

2023-06-23 11:02:19 +02:00

(Migrated from pm.bsc.es)

approved this merge request

rarias commented

2023-06-23 11:18:10 +02:00

(Migrated from pm.bsc.es)

I would say yes, as we want to have a trusted MPI program available to test the nodes at all times. Is very convenient to run srun -N2 osu_bw and know it will just work. It is 5.5MB in size, so space is not problematic.

I would say yes, as we want to have a trusted MPI program available to test the nodes at all times. Is very convenient to run `srun -N2 osu_bw` and know it will just work. It is 5.5MB in size, so space is not problematic.

rarias commented

2023-06-23 11:18:11 +02:00

(Migrated from pm.bsc.es)

resolved all threads