Overview
flytectl create registers new resources with FlyteAdmin.
flytectl create <resource> -p <project> -d <domain> [flags]
Subcommands
| Subcommand | Aliases | Description |
|---|
create project | projects | Register a new project |
create execution | executions | Launch a workflow or task execution |
create project
Register a new Flyte project.
# Create a project using flags
flytectl create project \
--name flytesnacks \
--id flytesnacks \
--description "flytesnacks description" \
--labels app=flyte
# Create a project from a YAML definition file
flytectl create project --file project.yaml
project.yaml):
id: "project-unique-id"
name: "Name"
labels:
values:
app: flyte
description: "Some description for the project."
The project name must not contain whitespace characters.
| Flag | Description |
|---|
--name | Display name of the project |
--id | Unique identifier for the project |
--description | Human-readable description |
--labels | Key=value labels (repeatable) |
--file | Path to a YAML project definition file |
--dryRun | Print the request without executing it |
create execution
Launch a workflow or task execution. The recommended approach is to first generate an execution spec file using flytectl get task or flytectl get launchplan, then pass that file to create execution.
Generate an execution spec file
From a task:flytectl get task -d development -p flytesnacks \
core.control_flow.merge_sort.merge --version v2 \
--execFile execution_spec.yaml
flytectl get launchplan -d development -p flytesnacks \
core.control_flow.merge_sort.merge_sort \
--execFile execution_spec.yaml
iamRoleARN: ""
inputs:
sorted_list1:
- 0
sorted_list2:
- 0
kubeServiceAcct: ""
targetDomain: ""
targetProject: ""
task: core.control_flow.merge_sort.merge
version: "v2"
Edit the inputs (optional)
Modify the spec file to supply real input values:iamRoleARN: 'arn:aws:iam::12345678:role/defaultrole'
inputs:
sorted_list1:
- 2
- 4
- 6
sorted_list2:
- 1
- 3
- 5
kubeServiceAcct: ""
targetDomain: ""
targetProject: ""
task: core.control_flow.merge_sort.merge
version: "v2"
targetExecutionCluster: "my-gpu-cluster"
Launch the execution
flytectl create execution \
--execFile execution_spec.yaml \
-p flytesnacks -d staging \
--targetProject flytesnacks
Relaunch an execution
flytectl create execution --relaunch ffb31066a0f8b4d52b77 \
-p flytesnacks -d development
Recover an execution
Recreate an execution from the last known failure point:
flytectl create execution --recover ffb31066a0f8b4d52b77 \
-p flytesnacks -d development
Named executions
Assign a custom name to an execution (must be unique within project+domain):
flytectl create execution --recover ffb31066a0f8b4d52b77 \
-p flytesnacks -d development my_custom_name
iamRoleARN: "arn:aws:iam::123456789:role/dummy"
inputs:
"x":
"x": 2
"y": ydatafory
"z":
1: "foo"
2: "bar"
"y":
"x": 3
"y": ydataforx
"z":
3: "buzz"
4: "lightyear"
kubeServiceAcct: ""
targetDomain: ""
targetProject: ""
task: core.type_system.custom_objects.add
version: v3
Assign to a cluster pool
flytectl create execution \
--execFile execution_spec.yaml \
-p flytesnacks -d development \
--clusterPool my-gpu-cluster
| Flag | Description |
|---|
--execFile | Path to the execution spec YAML file |
--relaunch | Execution ID to relaunch |
--recover | Execution ID to recover from last failure |
--targetProject | Project in which to create the execution |
--targetDomain | Domain in which to create the execution |
--targetExecutionCluster | Cluster to run the execution on |
--iamRoleARN | IAM role ARN for auth |
--kubeServiceAcct | Kubernetes service account for auth |
--version | Version of the workflow or task |
--clusterPool | Cluster pool to assign the execution to |
--overwriteCache | Skip cached results and recompute all outputs |
--dryRun | Print the request without executing it |
