From fae9cc7dd1f6ed07d2c52bd0df3bd958a539da52 Mon Sep 17 00:00:00 2001
From: Hugo Dominois <hugo.dominois@inria.fr>
Date: Tue, 25 Mar 2025 15:42:48 +0100
Subject: [PATCH 1/2] [sophia][esterel26] Add GPU device
---
.../sophia/clusters/esterel26/esterel26.json | 4 +-
.../clusters/esterel26/nodes/esterel26-1.json | 70 +++++++++++++++++++
.../clusters/esterel26/nodes/esterel26-1.yaml | 33 +++++++++
3 files changed, 105 insertions(+), 2 deletions(-)
diff --git a/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json b/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
index 733f18697a2..13491d7a144 100644
--- a/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
+++ b/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
@@ -53,8 +53,8 @@
],
"model": "Dell PowerEdge T640",
"nodes_count": 1,
- "nodes_description": "2 CPUs Intel Xeon Silver 4216, 16 cores/CPU, 384GB RAM, 558GB HDD, 3575GB HDD, 1 x 1Gb Ethernet, 1 x 40Gb InfiniBand",
- "priority": 202010,
+ "nodes_description": "2 CPUs Intel Xeon Silver 4216, 16 cores/CPU, 4 GPUs Quadro RTX 8000, 384GB RAM, 558GB HDD, 3575GB HDD, 1 x 1Gb Ethernet, 1 x 40Gb InfiniBand",
+ "priority": 202110,
"queues": [
"admin",
"testing"
diff --git a/data/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.json b/data/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.json
index fd4f22045de..3bf570c4722 100644
--- a/data/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.json
+++ b/data/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.json
@@ -20,6 +20,76 @@
"warranty_end": "2027-10-31"
},
"exotic": false,
+ "gpu_devices": {
+ "nvidia0": {
+ "compute_capability": "7.5",
+ "cores": 4608,
+ "cpu_affinity": 0,
+ "device": "/dev/nvidia0",
+ "memory": 48318382080,
+ "microarchitecture": "Turing",
+ "model": "Quadro RTX 8000",
+ "performance": {
+ "fp-16": 32620000000000,
+ "fp-32": 16310000000000,
+ "fp-64": 510000000000
+ },
+ "power_default_limit": "250.00 W",
+ "vbios_version": "90.02.4E.00.03",
+ "vendor": "Nvidia"
+ },
+ "nvidia1": {
+ "compute_capability": "7.5",
+ "cores": 4608,
+ "cpu_affinity": 0,
+ "device": "/dev/nvidia1",
+ "memory": 48318382080,
+ "microarchitecture": "Turing",
+ "model": "Quadro RTX 8000",
+ "performance": {
+ "fp-16": 32620000000000,
+ "fp-32": 16310000000000,
+ "fp-64": 510000000000
+ },
+ "power_default_limit": "250.00 W",
+ "vbios_version": "90.02.4E.00.03",
+ "vendor": "Nvidia"
+ },
+ "nvidia2": {
+ "compute_capability": "7.5",
+ "cores": 4608,
+ "cpu_affinity": 1,
+ "device": "/dev/nvidia2",
+ "memory": 48318382080,
+ "microarchitecture": "Turing",
+ "model": "Quadro RTX 8000",
+ "performance": {
+ "fp-16": 32620000000000,
+ "fp-32": 16310000000000,
+ "fp-64": 510000000000
+ },
+ "power_default_limit": "250.00 W",
+ "vbios_version": "90.02.4E.00.03",
+ "vendor": "Nvidia"
+ },
+ "nvidia3": {
+ "compute_capability": "7.5",
+ "cores": 4608,
+ "cpu_affinity": 1,
+ "device": "/dev/nvidia3",
+ "memory": 48318382080,
+ "microarchitecture": "Turing",
+ "model": "Quadro RTX 8000",
+ "performance": {
+ "fp-16": 32620000000000,
+ "fp-32": 16310000000000,
+ "fp-64": 510000000000
+ },
+ "power_default_limit": "250.00 W",
+ "vbios_version": "90.02.4E.00.03",
+ "vendor": "Nvidia"
+ }
+ },
"main_memory": {
"ram_size": 412316860416
},
diff --git a/input/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.yaml b/input/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.yaml
index 0d7d8daefab..c34ba13547a 100644
--- a/input/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.yaml
+++ b/input/grid5000/sites/sophia/clusters/esterel26/nodes/esterel26-1.yaml
@@ -16,6 +16,39 @@ esterel26-1:
manufacturer: Dell Inc.
name: PowerEdge T640
serial: 61GCH73
+ gpu_devices:
+ nvidia0:
+ cpu_affinity: 0
+ device: "/dev/nvidia0"
+ memory: 48318382080
+ model: Quadro RTX 8000
+ power_default_limit: 250.00 W
+ vbios_version: 90.02.4E.00.03
+ vendor: Nvidia
+ nvidia1:
+ cpu_affinity: 0
+ device: "/dev/nvidia1"
+ memory: 48318382080
+ model: Quadro RTX 8000
+ power_default_limit: 250.00 W
+ vbios_version: 90.02.4E.00.03
+ vendor: Nvidia
+ nvidia2:
+ cpu_affinity: 1
+ device: "/dev/nvidia2"
+ memory: 48318382080
+ model: Quadro RTX 8000
+ power_default_limit: 250.00 W
+ vbios_version: 90.02.4E.00.03
+ vendor: Nvidia
+ nvidia3:
+ cpu_affinity: 1
+ device: "/dev/nvidia3"
+ memory: 48318382080
+ model: Quadro RTX 8000
+ power_default_limit: 250.00 W
+ vbios_version: 90.02.4E.00.03
+ vendor: Nvidia
main_memory:
ram_size: 412316860416
memory_devices:
--
GitLab
From 89cff71acddc3991aff550df0a456b8da45ea5f7 Mon Sep 17 00:00:00 2001
From: Hugo Dominois <hugo.dominois@inria.fr>
Date: Tue, 25 Mar 2025 15:44:01 +0100
Subject: [PATCH 2/2] [sophia][esterel26] Add GPU metrics
---
.../sophia/clusters/esterel26/esterel26.json | 27 +++++++++++++++++++
.../clusters/esterel26/esterel26_metrics.yaml | 21 +++++++++++++++
2 files changed, 48 insertions(+)
diff --git a/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json b/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
index 13491d7a144..552a9c5d699 100644
--- a/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
+++ b/data/grid5000/sites/sophia/clusters/esterel26/esterel26.json
@@ -49,6 +49,33 @@
"port": 9100,
"protocol": "prometheus"
}
+ },
+ {
+ "description": "Default subset of metrics from Prometheus Nvidia DCGM Exporter",
+ "name": "prom_nvgpu_default_metrics",
+ "optional_period": 15000,
+ "period": 0,
+ "source": {
+ "id": [
+ "DCGM_FI_DEV_SM_CLOCK",
+ "DCGM_FI_DEV_MEM_CLOCK",
+ "DCGM_FI_DEV_GPU_TEMP",
+ "DCGM_FI_DEV_POWER_USAGE",
+ "DCGM_FI_DEV_MEM_COPY_UTIL"
+ ],
+ "port": 9400,
+ "protocol": "prometheus"
+ }
+ },
+ {
+ "description": "All metrics from Prometheus Nvidia DCGM Exporter",
+ "name": "prom_nvgpu_all_metrics",
+ "optional_period": 15000,
+ "period": 0,
+ "source": {
+ "port": 9400,
+ "protocol": "prometheus"
+ }
}
],
"model": "Dell PowerEdge T640",
diff --git a/input/grid5000/sites/sophia/clusters/esterel26/esterel26_metrics.yaml b/input/grid5000/sites/sophia/clusters/esterel26/esterel26_metrics.yaml
index 729e9beb582..120c97fe096 100644
--- a/input/grid5000/sites/sophia/clusters/esterel26/esterel26_metrics.yaml
+++ b/input/grid5000/sites/sophia/clusters/esterel26/esterel26_metrics.yaml
@@ -38,3 +38,24 @@ metrics:
source:
protocol: prometheus
port: 9100
+
+ - name: prom_nvgpu_default_metrics
+ description: Default subset of metrics from Prometheus Nvidia DCGM Exporter
+ period: 0
+ optional_period: 15000
+ source:
+ protocol: prometheus
+ port: 9400
+ id:
+ - DCGM_FI_DEV_SM_CLOCK
+ - DCGM_FI_DEV_MEM_CLOCK
+ - DCGM_FI_DEV_GPU_TEMP
+ - DCGM_FI_DEV_POWER_USAGE
+ - DCGM_FI_DEV_MEM_COPY_UTIL
+ - name: prom_nvgpu_all_metrics
+ description: All metrics from Prometheus Nvidia DCGM Exporter
+ period: 0
+ optional_period: 15000
+ source:
+ protocol: prometheus
+ port: 9400
--
GitLab