SambaRack Manager Node Administration - SambaNova Documentation

This guide covers viewing, managing, and performing DC power operations on SambaRack nodes.

Node commands

Use the following commands to view and manage nodes.

List nodes

Display all configured nodes and their types:

snctl node list

Example output - 4-PDU rack:

Configured Nodes
+-----------------+---------+---------+-------+
| Node Name       | Type    | Rack ID | Index |
+-----------------+---------+---------+-------+
| sn40l-16-001-n1 | 1001934 | 001     | 0     |
+-----------------+---------+---------+-------+

Example output - 2-PDU rack:

Configured Nodes
+-----------------+-----------+---------+-------+
| Node Name       | Type      | Rack ID | Index |
+-----------------+-----------+---------+-------+
| sn40l-16-001-n1 | 1002331 | 001     | 0     |
+-----------------+-----------+---------+-------+

Show node details

View detailed information about a specific node and its devices:

snctl node show <node_name>

Example:

snctl node show sn40l-16-001-n1

Example output:

Node Details: sn40l-16-001-n1
+-----------+--------------------+-------------+------+----------+
| Component | Hostname           | IP          | Port | Username |
+-----------+--------------------+-------------+------+----------+
| Host      | SFT-SN40L-16-H-1   | 172.20.3.45 | 3024 | root     |
| BMC       | SN40L-16-H-1-SP    | 172.20.3.46 |      | admin    |
| XRDU/0    | SN40L-16-H-1-XRDU0 | 172.20.3.47 | 3026 | root     |
| XRDU/1    | SN40L-16-H-1-XRDU1 | 172.20.3.48 | 3025 | root     |
| XRDU/2    | SN40L-16-H-1-XRDU2 | 172.20.3.49 | 3023 | root     |
| XRDU/3    | SN40L-16-H-1-XRDU3 | 172.20.3.50 | 3022 | root     |
| XRDU/4    | SN40L-16-H-1-XRDU4 | 172.20.3.51 | 3016 | root     |
| XRDU/5    | SN40L-16-H-1-XRDU5 | 172.20.3.52 | 3015 | root     |
| XRDU/6    | SN40L-16-H-1-XRDU6 | 172.20.3.53 | 3013 | root     |
| XRDU/7    | SN40L-16-H-1-XRDU7 | 172.20.3.54 | 3012 | root     |
+-----------+--------------------+-------------+------+----------+

Belongs to Rack: <rack_id> (Type: <rack_type>)

Rename node

Assign a custom name to a node:

snctl node rename <current_name> <new_name>

Example:

snctl node rename sn40l-16-001-n1 production-node-1

Device management (adding/removing devices) is done using snctl rack add-device and snctl rack delete-device. See Rack Administration for details.

Power operations

DC power operations control and monitor the power state of nodes and devices. These operations affect actual hardware. Supported devices:

Host systems: The CPU nodes in your infrastructure
XRDUs: The SambaNova execution units

Not supported:

Power distribution units (PDUs)
Network switches
Serial console servers
Other infrastructure components

Power operations perform DC cycle operations (power state control through host/XRDU management interfaces), not AC cycle operations (through PDUs).

Device types for power operations

The following device types are available for power operations:

Device type	Description
`node`	Entire node (host and all XRDUs)
`host`	Host system only
`xrdu`	XRDU devices only

Target selection

All power commands require either --node or --group (mutually exclusive):

Parameter	Description
`--node <node_id>`	Target a specific node
`--group <group_name>`	Target all nodes in a group

Check power state

Check the current power state of devices:

snctl node powerstate \
  (--node <node_id> | --group <group_name>) \
  --type <device_type>

Example – single node:

snctl node powerstate --node sn40l-16-001-n1 --type node

Example output:

Power State Information
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+
| Rack | Node Name       | Component   | Status | Host   | XRDU_0 | XRDU_1 | XRDU_2 | XRDU_3 | XRDU_4 | XRDU_5 | XRDU_6 | XRDU_7 |
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+
| 001  | sn40l-16-001-n1 | NODE_NODE_0 | PASS   | Online | Online | Online | Online | Online | Online | Online | Online | Online |
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+

Example – group:

snctl node powerstate --group g1 --type node

Example output:

Power State Information
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+
| Rack | Node Name       | Component   | Status | Host   | XRDU_0 | XRDU_1 | XRDU_2 | XRDU_3 | XRDU_4 | XRDU_5 | XRDU_6 | XRDU_7 |
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+
| 002  | sn40l-16-002-n1 | NODE_NODE_0 | PASS   | Online | Online | Online | Online | Online | Online | Online | Online | Online |
| 001  | sn40l-16-001-n1 | NODE_NODE_0 | PASS   | Online | Online | Online | Online | Online | Online | Online | Online | Online |
+------+-----------------+-------------+--------+--------+--------+--------+--------+--------+--------+--------+--------+--------+

Power on

Power on devices:

snctl node poweron \
  (--node <node_id> | --group <group_name>) \
  --type <device_type>

Example – single node:

snctl node poweron --node 001 --type node

Example output:

Node Power ON Status
└── Rack-001 🟢 Power ON Successful

Example – group:

snctl node poweron --group g1 --type node

Example output:

Node Power ON Status
└── Rack-001 🟢 Power ON Successful
└── Rack-002 🟢 Power ON Successful

Power off

Power off devices:

Power off operations interrupt running services. Plan accordingly.

snctl node poweroff \
  (--node <node_id> | --group <group_name>) \
  --type <device_type>

Example – single node:

snctl node poweroff --node 001 --type node

Example output:

Node Power OFF Status
└── Rack-001 🟢 Power OFF Successful

Example – group:

snctl node poweroff --group g1 --type node

Example output:

Node Power OFF Status
└── Rack-001 🟢 Power OFF Successful
└── Rack-002 🟢 Power OFF Successful

Power cycle

Perform a power cycle (off then on) on devices:

Power cycle operations interrupt running services. Plan accordingly.

snctl node powercycle \
  (--node <node_id> | --group <group_name>) \
  --type node

Power cycle currently only supports --type node.

Example – single node:

snctl node powercycle --node 001 --type node

Example output:

Node Power Cycle Status
└── Rack-001 🟢 Power Cycle Successful

Example – group:

snctl node powercycle --group g1 --type node

Example output:

Node Power Cycle Status
└── Rack-001 🟢 Power Cycle Successful
└── Rack-002 🟢 Power Cycle Successful

Using groups for batch operations

Groups allow you to perform power operations on multiple nodes at once:

# Check power state of all nodes in a group
snctl node powerstate --group production --type node

# Power on all nodes in a group
snctl node poweron --group production --type node

# Power off all nodes in a group
snctl node poweroff --group production --type node

# Power cycle all nodes in a group
snctl node powercycle --group production --type node

See Group Administration for information on creating and managing groups.

Getting Started

Service Administration

Hardware Administration

Reference Architecture

Resources

Node administration

Node commands

List nodes

Show node details

Rename node

Power operations

Device types for power operations

Target selection

Check power state

Power on

Power off

Power cycle

Using groups for batch operations

​Node commands

​List nodes

​Show node details

​Rename node

​Power operations

​Device types for power operations

​Target selection

​Check power state

​Power on

​Power off

​Power cycle

​Using groups for batch operations

Node commands

List nodes

Show node details

Rename node

Power operations

Device types for power operations

Target selection

Check power state

Power on

Power off

Power cycle

Using groups for batch operations