|
 |
1dc99f |
From f2cc67750f55a71edff0c527a1bfc14fde8132c3 Mon Sep 17 00:00:00 2001
|
|
|
1dc99f |
From: Mamatha Inamdar <mamatha4@linux.vnet.ibm.com>
|
|
|
1dc99f |
Date: Mon, 8 Nov 2021 10:50:03 +0530
|
|
|
1dc99f |
Subject: [PATCH] [nvidia]:Patch to update nvidia plugin for GPU info
|
|
|
1dc99f |
|
|
|
1dc99f |
This patch is to update nvidia plugin to collect
|
|
|
1dc99f |
logs for Nvidia GPUs
|
|
|
1dc99f |
|
|
|
1dc99f |
Signed-off-by: Mamatha Inamdar <mamatha4@linux.vnet.ibm.com>
|
|
|
1dc99f |
Reported-by: Borislav Stoymirski <borislav.stoymirski@bg.ibm.com>
|
|
|
1dc99f |
Reported-by: Yesenia Jimenez <yesenia@us.ibm.com>
|
|
|
1dc99f |
---
|
|
|
1dc99f |
sos/report/plugins/nvidia.py | 15 +++++++++++++--
|
|
|
1dc99f |
1 file changed, 13 insertions(+), 2 deletions(-)
|
|
|
1dc99f |
|
|
|
1dc99f |
diff --git a/sos/report/plugins/nvidia.py b/sos/report/plugins/nvidia.py
|
|
|
1dc99f |
index 09aaf586b..9e21b478e 100644
|
|
|
1dc99f |
--- a/sos/report/plugins/nvidia.py
|
|
|
1dc99f |
+++ b/sos/report/plugins/nvidia.py
|
|
|
1dc99f |
@@ -23,13 +23,24 @@ def setup(self):
|
|
|
1dc99f |
'--list-gpus',
|
|
|
1dc99f |
'-q -d PERFORMANCE',
|
|
|
1dc99f |
'-q -d SUPPORTED_CLOCKS',
|
|
|
1dc99f |
- '-q -d PAGE_RETIREMENT'
|
|
|
1dc99f |
+ '-q -d PAGE_RETIREMENT',
|
|
|
1dc99f |
+ '-q',
|
|
|
1dc99f |
+ '-q -d ECC',
|
|
|
1dc99f |
+ 'nvlink -s',
|
|
|
1dc99f |
+ 'nvlink -e'
|
|
|
1dc99f |
]
|
|
|
1dc99f |
|
|
|
1dc99f |
self.add_cmd_output(["nvidia-smi %s" % cmd for cmd in subcmds])
|
|
|
1dc99f |
|
|
|
1dc99f |
query = ('gpu_name,gpu_bus_id,vbios_version,temperature.gpu,'
|
|
|
1dc99f |
- 'utilization.gpu,memory.total,memory.free,memory.used')
|
|
|
1dc99f |
+ 'utilization.gpu,memory.total,memory.free,memory.used,'
|
|
|
1dc99f |
+ 'clocks.applications.graphics,clocks.applications.memory')
|
|
|
1dc99f |
+ querypages = ('timestamp,gpu_bus_id,gpu_serial,gpu_uuid,'
|
|
|
1dc99f |
+ 'retired_pages.address,retired_pages.cause')
|
|
|
1dc99f |
self.add_cmd_output("nvidia-smi --query-gpu=%s --format=csv" % query)
|
|
|
1dc99f |
+ self.add_cmd_output(
|
|
|
1dc99f |
+ "nvidia-smi --query-retired-pages=%s --format=csv" % querypages
|
|
|
1dc99f |
+ )
|
|
|
1dc99f |
+ self.add_journal(boot=0, identifier='nvidia-persistenced')
|
|
|
1dc99f |
|
|
|
1dc99f |
# vim: set et ts=4 sw=4 :
|