From be060aeabdcdc6f5e2e1b739d14993f83bb31cf6 Mon Sep 17 00:00:00 2001
From: Kerri Wait <kerri.wait@monash.edu>
Date: Tue, 6 Oct 2020 10:49:38 +1100
Subject: [PATCH] Add nvidia_smi plugin to telegraf config
---
.../telegraf/files/inputs.nvidia_smi.conf.j2 | 7 ++++
roles/telegraf/tasks/main.yml | 15 +++++++
roles/telegraf/templates/telegraf.conf.j2 | 42 +++++++++----------
3 files changed, 43 insertions(+), 21 deletions(-)
create mode 100644 roles/telegraf/files/inputs.nvidia_smi.conf.j2
diff --git a/roles/telegraf/files/inputs.nvidia_smi.conf.j2 b/roles/telegraf/files/inputs.nvidia_smi.conf.j2
new file mode 100644
index 00000000..79768fe7
--- /dev/null
+++ b/roles/telegraf/files/inputs.nvidia_smi.conf.j2
@@ -0,0 +1,7 @@
+# Pulls statistics from nvidia GPUs attached to the host
+[[inputs.nvidia_smi]]
+## Optional: path to nvidia-smi binary, defaults to $PATH via exec.LookPath
+# bin_path = "/usr/bin/nvidia-smi"
+
+## Optional: timeout for GPU polling
+# timeout = "5s"
\ No newline at end of file
diff --git a/roles/telegraf/tasks/main.yml b/roles/telegraf/tasks/main.yml
index f1950914..9226fa68 100644
--- a/roles/telegraf/tasks/main.yml
+++ b/roles/telegraf/tasks/main.yml
@@ -70,3 +70,18 @@
become_user: root
tags:
- configuration
+
+- name: Install nvidia-smi plugin
+ template:
+ src: inputs.nvidia_smi.conf.j2
+ dest: /etc/telegraf/telegraf.d/inputs.nvidia_smi.conf
+ owner: telegraf
+ group: telegraf
+ mode: '640'
+ notify:
+ - "restart telegraf"
+ become: true
+ become_user: root
+ tags:
+ - configuration
+ - gpu
\ No newline at end of file
diff --git a/roles/telegraf/templates/telegraf.conf.j2 b/roles/telegraf/templates/telegraf.conf.j2
index ae81c112..73338ecc 100644
--- a/roles/telegraf/templates/telegraf.conf.j2
+++ b/roles/telegraf/templates/telegraf.conf.j2
@@ -180,27 +180,27 @@
influxdb_database="slurm"
{% endif %}
-# Read mlx hardware counters
-{% if 'hw_counters' in ansible_local %}
-{% for interface in ansible_local['hw_counters'] %}
-[[inputs.multifile]]
- name_override = 'infiniband'
- base_dir = '/sys/class/infiniband'
- interval = '60s'
-
- [[inputs.multifile.tags]]
- device = '{{ interface }}'
- port = '1'
- type = 'hw_counters'
-
- {% for counter in ansible_local['hw_counters'][interface] | sort %}
-[[inputs.multifile.file]]
- file = '{{ interface }}/ports/1/hw_counters/{{ counter }}'
- conversion = 'int'
-
- {% endfor %}
-{% endfor %}
-{% endif %}
+## Read mlx hardware counters
+#{% if 'hw_counters' in ansible_local %}
+#{% for interface in ansible_local['hw_counters'] %}
+#[[inputs.multifile]]
+# name_override = 'infiniband'
+# base_dir = '/sys/class/infiniband'
+# interval = '60s'
+#
+# [[inputs.multifile.tags]]
+# device = '{{ interface }}'
+# port = '1'
+# type = 'hw_counters'
+#
+# {% for counter in ansible_local['hw_counters'][interface] | sort %}
+#[[inputs.multifile.file]]
+# file = '{{ interface }}/ports/1/hw_counters/{{ counter }}'
+# conversion = 'int'
+#
+# {% endfor %}
+#{% endfor %}
+#{% endif %}
{% if 'Karaage' in group_names %}
--
GitLab