Skip to content

Fabric Configuration

Overview

The fab.yaml file is the configuration file for the fabric. It supplies the configuration of the users, their credentials, logging, telemetry, and other non wiring related settings. The fab.yaml file is composed of multiple YAML objects inside of a single file. Per the YAML spec 3 hyphens (---) on a single line separate the end of one object from the beginning of the next. There are two YAML objects in the fab.yaml file. For more information about how to use hhfab init, run hhfab init --help.

HHFAB workflow

After hhfab has been downloaded:

  1. hhfab init(see different flags to customize initial configuration)
  2. Adjust the fab.yaml file to your needs
  3. Build your wiring diagram
  4. hhfab validate
  5. (optionally) hhfab diagram
  6. hhfab build

Or import existing fab.yaml and wiring files:

  1. hhfab init -c fab.yaml -w wiring-file.yaml -w extra-wiring-file.yaml
  2. hhfab validate
  3. Build your wiring diagram
  4. (optionally) hhfab diagram
  5. hhfab build

After the above workflow a user will have a .img file suitable for installing the control node, then bringing up the switches which comprise the fabric.

Complete Example File

The following example outlines a comprehensive Fabricator configuration. You can find further configuration details in the Fabricator API Reference.

fab.yaml
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
apiVersion: fabricator.githedgehog.com/v1beta1
kind: Fabricator
metadata:
  name: default
  namespace: fab
spec:
  config:
    control:
      tlsSAN: # IPs and DNS names to access API
        - "customer.site.io"

      ntpServers:
      - time.cloudflare.com
      - time1.google.com

      defaultUser: # username 'core' on all control nodes
        password: "hash..." # generate hash with openssl passwd -5
        authorizedKeys:
          - "ssh-ed25519 key..." # generate ssh key with ssh-keygen

    fabric:
      mode: spine-leaf # "spine-leaf" or "collapsed-core" (deprecated)
      includeONIE: true
      defaultSwitchUsers:
        admin: # at least one user with name 'admin' and role 'admin'
          role: admin
          password: "hash..." # generate hash with openssl passwd -5
          authorizedKeys:
            - "ssh-ed25519 key..."
        op: # optional read-only user
          role: operator
          password: "hash..." # generate hash with openssl passwd -5
          authorizedKeys:
            - "ssh-ed25519 key..." # generate ssh key with ssh-keygen

      defaultAlloyConfig:
        agentScrapeIntervalSeconds: 120
        unixScrapeIntervalSeconds: 120
        unixExporterEnabled: true
        collectSyslogEnabled: true
        lokiTargets:
          lab:
            url: http://url.io:3100/loki/api/v1/push
            labels:
              descriptive: name
        prometheusTargets:
          lab:
            url: http://url.io:9100/api/v1/push
            labels:
              descriptive: name
            sendIntervalSeconds: 120

---
apiVersion: fabricator.githedgehog.com/v1beta1
kind: ControlNode
metadata:
  name: control-1
  namespace: fab
spec:
  bootstrap:
    disk: "/dev/sda" # disk to install OS on, e.g. "sda" or "nvme0n1"
  external:
    interface: eno2 # customer interface to manage control node
    ip: dhcp # IP address for external interface
  management: # interface that manages switches in private management network
    interface: eno1

# Currently only one ControlNode is supported

Configure Control Node and Switch Users

Control Node Users

Configuring control node and switch users is done either passing --default-password-hash to hhfab init or editing the resulting fab.yaml file emitted by hhfab init. The default username on the control node is core.

Switch Users

There are two users on the switches, admin and operator. The operator user has read-only access to sonic-cli command on the switches. The admin user has broad administrative power on the switch. To avoid conflicts, do not use the following usernames: operator,hhagent,netops.

NTP and DHCP

The control node uses public NTP servers from Cloudflare and Google by default. The control node runs a DHCP server on the management network. See the example file.

Control Node

The control node is the host that manages all the switches, runs k3s, and serves images. The management interface is for the control node to manage the fabric switches, not end-user management of the control node. For end-user management of the control node specify the external interface name.

Telemetry

There is an option to enable Grafana Alloy on all switches to forward metrics and logs to the configured targets using Prometheus Remote-Write API and Loki API. Metrics includes port speeds, counters, errors, operational status, transceivers, fans, power supplies, temperature sensors, BGP neighbors, LLDP neighbors, and more. Logs include Hedgehog agent logs. Modify the URL as needed, instead of /api/v1/push it could be /api/v1/write; check the documentation for the data provider.

Switches push telemetry data through a proxy running in a pod on the control node. Switches do not have direct access to the Internet. Configure the control node to be able to reach and resolve the location of the Prometheus and Loki servers.

Telemetry can be enabled after installation of the fabric. Open the following YAML file in an editor on the control node. Modify the fields as needed. Logs can be pushed to a Grafana instance at the customer environment, or to Grafana cloud.

telemetry.yaml
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
spec:
  config:
    fabric:
      defaultAlloyConfig:
        agentScrapeIntervalSeconds: 120
        unixScrapeIntervalSeconds: 120
        unixExporterEnabled: true
        lokiTargets:
          grafana_cloud: # target name, multiple targets can be configured
              basicAuth: # optional
                  password: "<password>"
                  username: "<username>"
              labels: # labels to be added to all logs
                  env: env-1
              url: https://logs-prod-021.grafana.net/loki/api/v1/push
              useControlProxy: true
        prometheusTargets:
          grafana_cloud: # target name, multiple targets can be configured
              basicAuth: # optional
                  password: "<password>"
                  username: "<username>"
              labels: # labels to be added to all metrics
                  env: env-1
              sendIntervalSeconds: 120
              url: https://prometheus-prod-36-prod-us-west-0.grafana.net/api/prom/push
              useControlProxy: true
        unixExporterCollectors: # list of node-exporter collectors to enable, https://grafana.com/docs/alloy/latest/reference/components/prometheus.exporter.unix/#collectors-list
        - cpu
        - filesystem
        - loadavg
        - meminfo
        collectSyslogEnabled: true # collect /var/log/syslog on switches and forward to the lokiTargets

To enable the telemetry after install use:

kubectl patch -n fab --type merge fabricator/default --patch-file telemetry.yaml

For additional options, see the AlloyConfig struct in Fabric repo.