Hardware Observer

Learn about configurations >

• collect-timeout | int

  Default: 10

  Timeout for collectors' shell commands in seconds. Changing this will also change the scrape_timeout config option for prometheus for the cos-agent relation with observability agent like grafana agent or opentelemetry collector. This value is also used for the redfish client's timeout parameter. The value of this timeout should not be greater than prometheus scrape_interval (which is 60 seconds by default), as it greater would cause the scrape_timeout to be greater than scrape_interval.

• dcgm-snap-channel | string

  Default: auto

  Valid inputs are auto, v3/stable, v3/candidate, v3/edge, v4/stable, v4/candidate, v4/edge.

  When set to auto, the charm automatically checks the installed NVIDIA driver version and selects the most appropriate channel. For example, with driver 580, it will install from v4-cuda13/stable. The charm won't block if the driver is not installed or loaded.

  The v3 channels are compatible with CUDA 10, 11, and 12; you simply choose the desired release risk (stable, candidate, or edge).

  The v4 channels provide more flexibility. Even if you select v4/edge, the charm still detects the driver version and installs the matching CUDA build. For example, with driver 570, it will install from v4-cuda12/edge.

• exporter-log-level | string

  Default: INFO

  Start the prometheus exporter with log level "exporter-log-level". By default, it will set to INFO. Allowed values are "DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL". Values other than those will result in failure of the exporter.

• hardware-exporter-port | int

  Default: 10200

  Start the prometheus hardware exporter at "hardware-exporter-port". By default, it will start at port 10200.

• ipmi-driver-type | string

  Specify the driver type to be used for IPMI. Use LAN_2_0 for IPMI over LAN.

• redfish-disable | boolean

  Default: True

  By default the Redfish collector is disabled. Set this option to false to enable it.

• redfish-password | string

  This password is used to log in to the Redfish server and the BMC when IPMI over LAN is enabled.

• redfish-username | string

  This username is used to log in to the Redfish server and the BMC when IPMI over LAN is enabled.

• smartctl-exporter-port | int

  Default: 10201

  Start the prometheus smartctl exporter at "smartctl-exporter-port". By default, it will start at port 10201.

• smartctl-exporter-snap-channel | string

  Default: latest/stable

  Channel to install the Smartctl exporter snap if the hardware has smart disk. By default, it will install from latest/stable.