You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: design-proposals/separation-of-onboard-and-provisioning.md
+84-32Lines changed: 84 additions & 32 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -6,62 +6,112 @@ Last updated: 17th Nov 2025
6
6
7
7
## Abstract
8
8
9
-
Existing onboarding of edge node has to go through the provisioning workflow to install the operating system and required agents so that edge node is ready to manage by EMF orchestrator. It becomes to mandatory to test any day2 flows like VPro features, cluster orchestration, app orchestration. If customer or the user has an edge node which is already provisioned with the required Operation system meeting EMF requirements then it is an additional overhead to go through entire OS provisioning workflow to repeat the same process. Customer might have their own OS provisioning to install the OS in the edge node then we should support the onboarding of that edge node to EMF orchestrator and start using the day2 workflows. It becomes neccessity to split the onboarding and provisioning flow and it should be configurable by the end user when they want to onboard an edge to EMF to go through only onboarding flow without OS provisiong or with OS provisioning.
9
+
Existing onboarding of edge node has to go through the provisioning workflow
10
+
to install the operating system and required agents so that edge node is ready
11
+
to manage by EMF orchestrator. It becomes to mandatory to test any day2 flows
12
+
like VPro features, cluster orchestration, app orchestration. If customer or
13
+
the user has an edge node which is already provisioned with the required
14
+
Operation system meeting EMF requirements then it is an additional overhead to
15
+
go through entire OS provisioning workflow to repeat the same process. Customer
16
+
might have their own OS provisioning to install the OS in the edge node then we
17
+
should support the onboarding of that edge node to EMF orchestrator and start
18
+
using the day2 workflows. It becomes neccessity to split the onboarding and
19
+
provisioning flow and it should be configurable by the end user when they want
20
+
to onboard an edge to EMF to go through only onboarding flow without OS
21
+
provisiong or with OS provisioning.
10
22
11
23
## Requirements
12
24
13
-
When an user choose to onboard an edge node to EMF orchestrator, there should be an option to use to opt for provisioning flow or skip it entirely. When user opt for the onboard with OS provisioning option then it is exising workflow to go through the entire the Day0 workflow(installation OS, edge node installation and configuration). OS provisioning workflow differs between the 2 EMF supported Operating systems(EMT and Ubuntu) because EMT is pre-bundled with Edge node agents because immutability. Existing provsioning workflow for Ubuntu includes below steps
25
+
When an user choose to onboard an edge node to EMF orchestrator, there should
26
+
be an option to use to opt for provisioning flow or skip it entirely. When user
27
+
opt for the onboard with OS provisioning option then it is exising workflow to
28
+
go through the entire the Day0 workflow(installation OS, edge node installation
29
+
and configuration). OS provisioning workflow differs between the 2 EMF
30
+
supported Operating systems(EMT and Ubuntu) because EMT is pre-bundled with
31
+
Edge node agents because immutability.
32
+
33
+
### Existing provsioning workflow for Ubuntu includes below steps
14
34
15
35
1. Ubuntu installation by downloading the canonical base server image
16
36
17
-
2. Upgrade the required base kernel as part provisioning flow to support intel platforms features. Required base kernel for Intel platforms like ADL, RPL are
37
+
2. Upgrade the required base kernel as part provisioning flow to support intel
38
+
platforms features. Required base kernel for Intel platforms like ADL, RPL are
18
39
- Ubuntu 22.04 -> Kernel 6.8.x
19
40
20
41
- Ubuntu 24.05 -> Kernel 6.11.x
21
-
22
-
3. Install edge node agents as post installation of OS. DKAM curates the installer script with required EMF compatible versions of EN agents(by making use of EN manifest file) along with their configurations. It configures apt package manager with EMF release service where edge node agents debians are hosted.
23
-
24
-
4. Edge node agent configurations include resource managers end points, key cloak credentials, Edge node proxy configurations, edge node agent's log rotation policies, etc.
25
-
26
-
5. Starting the edge node agent as systemd services and enable them to start reboot.
27
42
43
+
3. Install edge node agents as post installation of OS. DKAM curates the
44
+
installer script with required EMF compatible versions of EN agents
45
+
(by making use of EN manifest file) along with their configurations.
46
+
It configures apt package manager with EMF release service where edge node
47
+
agents debians are hosted.
28
48
29
-
EMF orchstrator side steps to skip the provisioning flow
49
+
4. Edge node agent configurations include resource managers end points,
1. User should be able to register the edge node with EMF orchestrator by opting or skipping the provisioning flow.
53
+
5. Starting the edge node agent as systemd services and enable them to start
54
+
reboot.
32
55
56
+
### EMF orchstrator side steps to skip the provisioning flow
33
57
34
-
If user opt for the onboarding without OS provisioning then user should minimal steps in the edge node to onboard to EMF orchestrator.
35
-
Onboarding steps those should be done on the edge node includes
58
+
1. Users should be able to register edge nodes with the EMF orchestrator by
59
+
opting in or skipping the provisioning flow. When users choose onboarding
60
+
without OS provisioning, they must complete minimal steps on the edge node to
61
+
register with the EMF orchestrator. The onboarding steps that must be performed
62
+
on the edge node include:
36
63
37
-
1. Kernel upgrade to installed base kernel version required for the intel supported platforms as defined above provisioning flow.
64
+
- Kernel upgrade to installed base kernel version required for the intel
65
+
supported platforms as defined above provisioning flow.
38
66
39
-
2. Edge node agents installation
67
+
- Edge node agents installation
40
68
41
-
- Additional system packages to be installed on the edge node to run the edge node agents.
69
+
- Additional system packages to be installed on the edge node to run the
70
+
edge node agents.
42
71
43
-
- It includes Edge node agents configuration files with all required version of EN agents which is compatible with EMF, infra-managers end points, keycloak credentials etc. This configuration also include what are specific agents to be installed on the edge node based on EMF capabilities(EIM with AMT, App orchestration, cluster orchestartion, observability)
72
+
- It includes Edge node agents configuration files with all required
73
+
version of EN agents which is compatible with EMF, infra-managers end-
74
+
points, keycloak credentials etc. This configuration also include what are
75
+
specific agents to be installed on the edge node based on EMF capabilities
76
+
(EIM with AMT, App orchestration, cluster orchestartion, observability)
44
77
45
-
3. Run the new onboarding agent to perform Non-interactive onboarding and get the keycloak host specific credentials for the edge node agents.
78
+
2. Run the new onboarding agent to perform Non-interactive onboarding and get
79
+
the keycloak host specific credentials for the edge node agents.
46
80
47
-
4. Start other agents once key cloak credentials received from EMF orchestrator and enable them to start on reboot.
81
+
3. Start other agents once key cloak credentials received from EMF orchestrator
82
+
and enable them to start on reboot.
48
83
49
-
5. If the host resource is associated with custom config(cloud-init) then it there should be a way to run run cloud-init stpes on the edge node as a post onboarding step.
84
+
4. If the host resource is associated with custom config(cloud-init) then it
85
+
there should be a way to run run cloud-init stpes on the edge node as a post
86
+
onboarding step.
50
87
51
88
## Scope and Implementation plan
52
89
53
90
High level tasks in EMF to make the provisioning workflow as optional
54
91
55
-
1. Device disocery agent - Build debian for the device discovery agent and include it in the agent installer script which does onboarding(non-interactive) and gets required keycloak credentials to the edge node. Until device discovery completes the onboarding other agents installation shouldn't be started.
92
+
1. Device disocery agent - Build debian for the device discovery agent and
93
+
include it in the agent installer script which does onboarding(non-interactive)
94
+
and gets required keycloak credentials to the edge node. Until device discovery
95
+
completes the onboarding other agents installation shouldn't be started.
56
96
57
-
2. DKAM - should curate and host the installer script in the tinker-nginx service. It should also read the configuration with enabled capabilities of EMF (App orchestration, cluster orchestration, observability). Based on that configuration it shall include the respective agent installations. It should include device discovery agent.
97
+
2. DKAM - should curate and host the installer script in the tinker-nginx
98
+
service. It should also read the configuration with enabled capabilities of
99
+
EMF (App orchestration, cluster orchestration, observability). Based on that
100
+
configuration it shall include the respective agent installations. It should
101
+
include device discovery agent.
58
102
59
-
3. Onboarding manager - Should skip the provisioning flow to creation of the tinkerbell workflow if edge node is registered with skip provisioning flow option. It should update the inventory with required instance resource fields like provisioning status and status indicator fields. Creation of instance resource with mapping OS resource of Edge node.
103
+
3. Onboarding manager - Should skip the provisioning flow to creation of the
104
+
tinkerbell workflow if edge node is registered with skip provisioning flow
105
+
option. It should update the inventory with required instance resource fields
106
+
like provisioning status and status indicator fields. Creation of instance
107
+
resource with mapping OS resource of Edge node.
60
108
61
-
4. API-v2 and inventory changes to include new field, skip provisioning flow in host resource.
62
-
63
-
5. Orch-cli/Infra web-ui changes for device registration to include the new field skip provisiong flow. By default skip provisioning flow will be set to false.
109
+
4. API-v2 and inventory changes to include new field, skip provisioning flow in
110
+
host resource.
64
111
112
+
5. Orch-cli/Infra web-ui changes for device registration to include the new
113
+
field skip provisiong flow. By default skip provisioning flow will be set to
114
+
false.
65
115
66
116
## Workflow
67
117
@@ -91,7 +141,7 @@ autonumber
91
141
92
142
93
143
Note over User,EdgeNode: User shall trigger the onboarding flow
94
-
User->>EdgeNode: Login to edge node which has Ubuntu 22.04 or 24.04 pre-installed
144
+
User->>EdgeNode: Login to edge node which has Ubuntu 22.04 or 24.04 pre-installed
95
145
EdgeNode->>TinkerNginx: Download installer script to edge node
96
146
EdgeNode->>EdgeNode: Run the installer script Install system packages
EdgeNode->>EdgeNode: Install node agent and other EN agents
107
157
EdgeNode->>EdgeNode: Configure agents with onboading credentials
108
-
EdgeNode->>EdgeNode: Edge agent communicate with respective Infra managers using JWT token
158
+
EdgeNode->>EdgeNode: EN agent communicate with respective Infra managers
109
159
EdgeNode->>EdgeNode: Enable and Start all agents as systemd services
110
160
end
111
161
EdgeNode->>EdgeNode: Ready for Day2 operations(Update & remote power management)
112
162
```
113
163
114
164
## Opens
115
165
116
-
- In the interactive onboarding mapping the instance resource to OS resource is done by taking OS version from the Edge node during device discovery stage.
117
-
118
-
- Cluster creation might need the required partitions to be created which is done during the provisioning flow.
166
+
- In the interactive onboarding mapping the instance resource to OS resource
167
+
is done by taking OS version from the Edge node during device discovery stage.
119
168
120
-
- Kernel might need to be upgraded based on platform in order to use the platform features.
169
+
- Cluster creation might need the required partitions to be created which is
170
+
done during the provisioning flow.
121
171
172
+
- The kernel may need to be upgraded depending on the platform to enable
0 commit comments