nvidia-dcgm-exporter pod keeps crashing about dcgm-exporter

[nikkon-dev]

Hi,

You need to check if dcgm-exporter is able to connect to the remove nv-hostengine DCGM_REMOTE_HOSTENGINE_INFO: $(NODE_IP):5555.

There are two modes dcgm-exporter supports: 1) dcgm-exporter connects to a standalone nv-hostengine working in another container or right on the node. 2) dcgm-exporter runs embedded hostengine and does not need to connect to any other hostengine on port 5555. In the second mode, dcgm-exporter would fail to initialize some modules if there is another nv-hostengine process on the same physical machine.

You need to see the dcgm-exporter command line to understand the mode it's running.

[soumen-ghosh]

Hi,

You need to see the dcgm-exporter command line to understand the mode it's running.
Can you help with some pointers on how do I check this? I'm trying this for the 1st time.

[nikkon-dev]

Please, take a look here

dcgm-exporter/deployment/values.yaml

Line 23 in 0869a2a

arguments: ["-f", "/etc/dcgm-exporter/dcp-metrics-included.csv"]