BOS Options

NOTE This section is for Boot Orchestration Service (BOS) v2 only.

BOS provides a global service options endpoint for modifying the base configuration of the service itself. These options are available only for the BOS v2 API and only affect v2 functionality.

View options

(ncn-mw#) View the current option values with the following command:

cray bos v2 options list --format json

Example output:

{
  "cleanup_completed_session_ttl": "7d",
  "clear_stage": false,
  "component_actual_state_ttl": "4h",
  "default_retry_policy": 3,
  "disable_components_on_completion": true,
  "discovery_frequency": 300,
  "logging_level": "INFO",
  "max_boot_wait_time": 600,
  "max_power_off_wait_time": 180,
  "max_power_on_wait_time": 30,
  "polling_frequency": 60
}

Update options

(ncn-mw#) The values for all BOS global options can be modified with the cray bos v2 options update command. For example:

cray bos v2 options update --polling-frequency 12 --format json

Example output:

{
  "cleanup_completed_session_ttl": "7d",
  "clear_stage": false,
  "component_actual_state_ttl": "4h",
  "default_retry_policy": 3,
  "disable_components_on_completion": true,
  "discovery_frequency": 300,
  "logging_level": "INFO",
  "max_boot_wait_time": 600,
  "max_power_off_wait_time": 180,
  "max_power_on_wait_time": 30,
  "polling_frequency": 12
}

Individual option details

cleanup_completed_session_ttl

The amount of time that a completed BOS session can exist without being cleaned up by the session-cleanup operator.

The value can either be 0 or else be a non-negative integer following by a character indicating the units: minutes (mor M), hours (h or H), days (d or D), or weeks (w or W). For example, 3d means three days.

The cleanup behavior is disabled if the option is set to 0, 0m, 0h, 0d, or 0w.

clear_stage

Allows staged information for BOS components to be cleared when the requested staging action has been started. Defaults to false.

For more information on staging, see Stage Changes with BOS.

component_actual_state_ttl

This option defines two things:

  • The amount of time that a component’s actual_state is considered valid; if the actual state was last updated longer ago than this time, then the actual-state-cleanup operator will clear the actual state of the component.
  • 4/3 (133%) of the amount of time that the BOS reporter waits between sending state updates to BOS.
    • The BOS reporter does not read the BOS options during its execution, so changes to this option will not be reflected in the behavior of the BOS reporter on booted nodes.
    • For more details, see Reporting interval.

The value can either be 0 or else be a non-negative integer following by a character indicating the units: minutes (mor M), hours (h or H), days (d or D), or weeks (w or W). For example, 3d means three days.

WARNING: Unlike cleanup_completed_session_ttl, a zero value for this option will not disable the cleanup behavior; instead it will result in undesirable behavior. Specifically, component actual states will be cleared every time the actual-state-cleanup operator runs, and the BOS reporter will have no pauses between reporting the component status to BOS. This will effectively render BOS unable to properly manage the nodes.

To avoid problems, never set this option to a value less than 1 hour.

default_retry_policy

The default maximum number of attempts per node for failed actions.

disable_components_on_completion

This is an experimental feature and is not fully supported. This option is removed in CSM 1.7.

Determines if a component will be marked as disabled after its desired state matches its current state. If false, BOS will continue to maintain the state of the nodes declaratively.

discovery_frequency

The frequency with which BOS checks the Hardware State Manager (HSM) for new components and adds them to the BOS component database.

logging_level

The logging level for the BOS API server and the BOS Operators. Valid values for this option are DEBUG, INFO, and WARN.

max_boot_wait_time

How long BOS will wait for a node to boot into a usable state before rebooting it again (in seconds).

max_power_off_wait_time

How long BOS will wait for a node to power off before forcefully powering it off (in seconds).

max_power_on_wait_time

How long BOS will wait for a node to power on before calling power on again (in seconds).

polling_frequency

How frequently the BOS operators check component state for needed actions (in seconds).