04f99cc1c4
Add 6 application endpoints and schemas to OpenAPI spec. Update .gitignore. Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent) Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
1796 lines
50 KiB
JSON
1796 lines
50 KiB
JSON
{
|
|
"openapi": "3.0.3",
|
|
"info": {
|
|
"title": "GCY HPC Server API",
|
|
"description": "Slurm HPC cluster management API for job submission and cluster monitoring.",
|
|
"version": "0.1.0"
|
|
},
|
|
"servers": [
|
|
{
|
|
"url": "/api/v1",
|
|
"description": "API v1"
|
|
}
|
|
],
|
|
"tags": [
|
|
{
|
|
"name": "Jobs",
|
|
"description": "Job submission, listing, cancellation, and history"
|
|
},
|
|
{
|
|
"name": "Cluster",
|
|
"description": "Cluster node, partition, and diagnostic information"
|
|
},
|
|
{
|
|
"name": "Applications",
|
|
"description": "Application definition CRUD and job submission"
|
|
}
|
|
],
|
|
"paths": {
|
|
"/jobs/submit": {
|
|
"post": {
|
|
"tags": ["Jobs"],
|
|
"summary": "Submit a new job",
|
|
"description": "Submits a Slurm job with the specified script and optional parameters.",
|
|
"operationId": "submitJob",
|
|
"requestBody": {
|
|
"required": true,
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/SubmitJobRequest"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"responses": {
|
|
"201": {
|
|
"description": "Job submitted successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/JobResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid request body or missing required fields",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
},
|
|
"examples": {
|
|
"invalid_body": {
|
|
"value": {
|
|
"success": false,
|
|
"error": "invalid request body"
|
|
}
|
|
},
|
|
"missing_script": {
|
|
"value": {
|
|
"success": false,
|
|
"error": "script is required"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"502": {
|
|
"description": "Slurm backend error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/jobs": {
|
|
"get": {
|
|
"tags": ["Jobs"],
|
|
"summary": "List all jobs with pagination",
|
|
"description": "Retrieves a paginated list of all current jobs.",
|
|
"operationId": "getJobs",
|
|
"parameters": [
|
|
{
|
|
"name": "page",
|
|
"in": "query",
|
|
"description": "Page number (starts from 1)",
|
|
"required": false,
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 1
|
|
}
|
|
},
|
|
{
|
|
"name": "page_size",
|
|
"in": "query",
|
|
"description": "Number of items per page",
|
|
"required": false,
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 20
|
|
}
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Paginated list of jobs",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/JobListResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid query parameters",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/jobs/history": {
|
|
"get": {
|
|
"tags": ["Jobs"],
|
|
"summary": "Get job history",
|
|
"description": "Retrieves paginated job history with optional filters.",
|
|
"operationId": "getJobHistory",
|
|
"parameters": [
|
|
{
|
|
"name": "users",
|
|
"in": "query",
|
|
"description": "Filter by username(s)",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "start_time",
|
|
"in": "query",
|
|
"description": "Filter by start time (Unix timestamp or date string)",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "end_time",
|
|
"in": "query",
|
|
"description": "Filter by end time (Unix timestamp or date string)",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "account",
|
|
"in": "query",
|
|
"description": "Filter by account",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "partition",
|
|
"in": "query",
|
|
"description": "Filter by partition",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "state",
|
|
"in": "query",
|
|
"description": "Filter by job state",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "job_name",
|
|
"in": "query",
|
|
"description": "Filter by job name",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "submit_time",
|
|
"in": "query",
|
|
"description": "Filter by submit time (Unix timestamp)",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "cluster",
|
|
"in": "query",
|
|
"description": "Filter by cluster name",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "qos",
|
|
"in": "query",
|
|
"description": "Filter by QOS policy",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "constraints",
|
|
"in": "query",
|
|
"description": "Filter by node constraints",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "exit_code",
|
|
"in": "query",
|
|
"description": "Filter by exit code",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "node",
|
|
"in": "query",
|
|
"description": "Filter by allocated node",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "reservation",
|
|
"in": "query",
|
|
"description": "Filter by reservation name",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "groups",
|
|
"in": "query",
|
|
"description": "Filter by user groups",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "wckey",
|
|
"in": "query",
|
|
"description": "Filter by WCKey (Workload Characterization Key)",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
{
|
|
"name": "page",
|
|
"in": "query",
|
|
"description": "Page number (default: 1)",
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 1
|
|
}
|
|
},
|
|
{
|
|
"name": "page_size",
|
|
"in": "query",
|
|
"description": "Items per page (default: 20)",
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 20
|
|
}
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Paginated job history",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/JobListResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid query parameters",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/jobs/{id}": {
|
|
"get": {
|
|
"tags": ["Jobs"],
|
|
"summary": "Get job by ID",
|
|
"description": "Retrieves details for a specific job. If the job is not found in the active queue, automatically queries SlurmDBD history.",
|
|
"operationId": "getJob",
|
|
"parameters": [
|
|
{
|
|
"$ref": "#/components/parameters/JobId"
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Job details",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/JobResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Job not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"delete": {
|
|
"tags": ["Jobs"],
|
|
"summary": "Cancel a job",
|
|
"description": "Cancels a running or pending job by ID.",
|
|
"operationId": "cancelJob",
|
|
"parameters": [
|
|
{
|
|
"$ref": "#/components/parameters/JobId"
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Job cancelled successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "object",
|
|
"properties": {
|
|
"message": {
|
|
"type": "string",
|
|
"example": "job cancelled"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"502": {
|
|
"description": "Slurm backend error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/nodes": {
|
|
"get": {
|
|
"tags": ["Cluster"],
|
|
"summary": "List all nodes",
|
|
"description": "Retrieves a list of all cluster nodes.",
|
|
"operationId": "getNodes",
|
|
"responses": {
|
|
"200": {
|
|
"description": "List of nodes",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/NodeResponse"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/nodes/{name}": {
|
|
"get": {
|
|
"tags": ["Cluster"],
|
|
"summary": "Get node by name",
|
|
"description": "Retrieves details for a specific cluster node.",
|
|
"operationId": "getNode",
|
|
"parameters": [
|
|
{
|
|
"$ref": "#/components/parameters/NodeName"
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Node details",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/NodeResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Node not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/partitions": {
|
|
"get": {
|
|
"tags": ["Cluster"],
|
|
"summary": "List all partitions",
|
|
"description": "Retrieves a list of all cluster partitions.",
|
|
"operationId": "getPartitions",
|
|
"responses": {
|
|
"200": {
|
|
"description": "List of partitions",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/PartitionResponse"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/partitions/{name}": {
|
|
"get": {
|
|
"tags": ["Cluster"],
|
|
"summary": "Get partition by name",
|
|
"description": "Retrieves details for a specific cluster partition.",
|
|
"operationId": "getPartition",
|
|
"parameters": [
|
|
{
|
|
"$ref": "#/components/parameters/PartitionName"
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Partition details",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/PartitionResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Partition not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/applications": {
|
|
"get": {
|
|
"tags": ["Applications"],
|
|
"summary": "List applications",
|
|
"description": "Retrieves a paginated list of application definitions.",
|
|
"operationId": "listApplications",
|
|
"parameters": [
|
|
{
|
|
"name": "page",
|
|
"in": "query",
|
|
"description": "Page number (starts from 1)",
|
|
"required": false,
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 1
|
|
}
|
|
},
|
|
{
|
|
"name": "page_size",
|
|
"in": "query",
|
|
"description": "Number of items per page",
|
|
"required": false,
|
|
"schema": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"default": 20
|
|
}
|
|
}
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Paginated list of applications",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/ApplicationListResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"post": {
|
|
"tags": ["Applications"],
|
|
"summary": "Create an application",
|
|
"description": "Creates a new application definition.",
|
|
"operationId": "createApplication",
|
|
"requestBody": {
|
|
"required": true,
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/CreateApplicationRequest"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"responses": {
|
|
"201": {
|
|
"description": "Application created successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "object",
|
|
"properties": {
|
|
"id": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Created application ID"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid request body or missing required fields",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/applications/{id}": {
|
|
"get": {
|
|
"tags": ["Applications"],
|
|
"summary": "Get application by ID",
|
|
"description": "Retrieves a specific application definition.",
|
|
"operationId": "getApplication",
|
|
"parameters": [
|
|
{ "$ref": "#/components/parameters/ApplicationId" }
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Application details",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/Application"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Application not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"put": {
|
|
"tags": ["Applications"],
|
|
"summary": "Update an application",
|
|
"description": "Updates an existing application definition.",
|
|
"operationId": "updateApplication",
|
|
"parameters": [
|
|
{ "$ref": "#/components/parameters/ApplicationId" }
|
|
],
|
|
"requestBody": {
|
|
"required": true,
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/UpdateApplicationRequest"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"responses": {
|
|
"200": {
|
|
"description": "Application updated successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "object",
|
|
"properties": {
|
|
"message": {
|
|
"type": "string",
|
|
"example": "application updated"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid request body",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Application not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"delete": {
|
|
"tags": ["Applications"],
|
|
"summary": "Delete an application",
|
|
"description": "Deletes an application definition by ID.",
|
|
"operationId": "deleteApplication",
|
|
"parameters": [
|
|
{ "$ref": "#/components/parameters/ApplicationId" }
|
|
],
|
|
"responses": {
|
|
"200": {
|
|
"description": "Application deleted successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "object",
|
|
"properties": {
|
|
"message": {
|
|
"type": "string",
|
|
"example": "application deleted"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Application not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/applications/{id}/submit": {
|
|
"post": {
|
|
"tags": ["Applications"],
|
|
"summary": "Submit a job from an application",
|
|
"description": "Submits a Slurm job using the application's script template and provided parameter values.",
|
|
"operationId": "submitApplication",
|
|
"parameters": [
|
|
{ "$ref": "#/components/parameters/ApplicationId" }
|
|
],
|
|
"requestBody": {
|
|
"required": true,
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApplicationSubmitRequest"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"responses": {
|
|
"201": {
|
|
"description": "Job submitted successfully",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"$ref": "#/components/schemas/JobResponse"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"400": {
|
|
"description": "Invalid request body or missing required parameter values",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"404": {
|
|
"description": "Application not found",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"502": {
|
|
"description": "Slurm backend error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"/diag": {
|
|
"get": {
|
|
"tags": ["Cluster"],
|
|
"summary": "Get cluster diagnostics",
|
|
"description": "Retrieves diagnostic information about the Slurm cluster.",
|
|
"operationId": "getDiag",
|
|
"responses": {
|
|
"200": {
|
|
"description": "Cluster diagnostics",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"allOf": [
|
|
{ "$ref": "#/components/schemas/ApiResponseSuccess" },
|
|
{
|
|
"type": "object",
|
|
"properties": {
|
|
"data": {
|
|
"type": "object",
|
|
"description": "Slurm cluster diagnostic data"
|
|
}
|
|
}
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"500": {
|
|
"description": "Internal server error",
|
|
"content": {
|
|
"application/json": {
|
|
"schema": {
|
|
"$ref": "#/components/schemas/ApiResponseError"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"components": {
|
|
"parameters": {
|
|
"JobId": {
|
|
"name": "id",
|
|
"in": "path",
|
|
"required": true,
|
|
"description": "Job ID",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
"NodeName": {
|
|
"name": "name",
|
|
"in": "path",
|
|
"required": true,
|
|
"description": "Node name",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
"PartitionName": {
|
|
"name": "name",
|
|
"in": "path",
|
|
"required": true,
|
|
"description": "Partition name",
|
|
"schema": {
|
|
"type": "string"
|
|
}
|
|
},
|
|
"ApplicationId": {
|
|
"name": "id",
|
|
"in": "path",
|
|
"required": true,
|
|
"description": "Application ID",
|
|
"schema": {
|
|
"type": "integer",
|
|
"format": "int64"
|
|
}
|
|
}
|
|
},
|
|
"schemas": {
|
|
"ApiResponseSuccess": {
|
|
"type": "object",
|
|
"required": ["success"],
|
|
"properties": {
|
|
"success": {
|
|
"type": "boolean",
|
|
"enum": [true]
|
|
},
|
|
"data": {
|
|
"description": "Response payload"
|
|
}
|
|
}
|
|
},
|
|
"ApiResponseError": {
|
|
"type": "object",
|
|
"required": ["success", "error"],
|
|
"properties": {
|
|
"success": {
|
|
"type": "boolean",
|
|
"enum": [false]
|
|
},
|
|
"error": {
|
|
"type": "string",
|
|
"description": "Error message"
|
|
}
|
|
}
|
|
},
|
|
"SubmitJobRequest": {
|
|
"type": "object",
|
|
"required": ["script"],
|
|
"properties": {
|
|
"script": {
|
|
"type": "string",
|
|
"description": "Job script content"
|
|
},
|
|
"partition": {
|
|
"type": "string",
|
|
"description": "Target partition"
|
|
},
|
|
"qos": {
|
|
"type": "string",
|
|
"description": "Quality of Service"
|
|
},
|
|
"cpus": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Number of CPUs required"
|
|
},
|
|
"memory": {
|
|
"type": "string",
|
|
"description": "Memory requirement (e.g. \"4G\")"
|
|
},
|
|
"time_limit": {
|
|
"type": "string",
|
|
"description": "Time limit (e.g. \"1:00:00\")"
|
|
},
|
|
"job_name": {
|
|
"type": "string",
|
|
"description": "Job name"
|
|
},
|
|
"environment": {
|
|
"type": "object",
|
|
"additionalProperties": {
|
|
"type": "string"
|
|
},
|
|
"description": "Environment variables"
|
|
}
|
|
}
|
|
},
|
|
"JobResponse": {
|
|
"type": "object",
|
|
"properties": {
|
|
"job_id": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Slurm job ID"
|
|
},
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Job name"
|
|
},
|
|
"job_state": {
|
|
"type": "array",
|
|
"items": {
|
|
"type": "string"
|
|
},
|
|
"description": "Current job state(s) (e.g. [\"RUNNING\"], [\"PENDING\",\"REQUEUED\"])"
|
|
},
|
|
"state_reason": {
|
|
"type": "string",
|
|
"description": "Reason for job pending or failure"
|
|
},
|
|
"partition": {
|
|
"type": "string",
|
|
"description": "Assigned partition"
|
|
},
|
|
"qos": {
|
|
"type": "string",
|
|
"description": "Quality of Service policy"
|
|
},
|
|
"priority": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Job priority",
|
|
"nullable": true
|
|
},
|
|
"time_limit": {
|
|
"type": "string",
|
|
"description": "Time limit in minutes (\"UNLIMITED\" for unlimited)"
|
|
},
|
|
"account": {
|
|
"type": "string",
|
|
"description": "Billing account"
|
|
},
|
|
"user": {
|
|
"type": "string",
|
|
"description": "Submitting user"
|
|
},
|
|
"cluster": {
|
|
"type": "string",
|
|
"description": "Cluster name"
|
|
},
|
|
"cpus": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Allocated/requested CPU cores",
|
|
"nullable": true
|
|
},
|
|
"tasks": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Number of tasks",
|
|
"nullable": true
|
|
},
|
|
"node_count": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Number of allocated nodes",
|
|
"nullable": true
|
|
},
|
|
"nodes": {
|
|
"type": "string",
|
|
"description": "Allocated node list"
|
|
},
|
|
"batch_host": {
|
|
"type": "string",
|
|
"description": "Batch host node"
|
|
},
|
|
"submit_time": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Submit time (Unix timestamp)",
|
|
"nullable": true
|
|
},
|
|
"start_time": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Start time (Unix timestamp)",
|
|
"nullable": true
|
|
},
|
|
"end_time": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "End time (Unix timestamp)",
|
|
"nullable": true
|
|
},
|
|
"exit_code": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Exit code (null if not finished)",
|
|
"nullable": true
|
|
},
|
|
"standard_output": {
|
|
"type": "string",
|
|
"description": "Stdout file path"
|
|
},
|
|
"standard_error": {
|
|
"type": "string",
|
|
"description": "Stderr file path"
|
|
},
|
|
"standard_input": {
|
|
"type": "string",
|
|
"description": "Stdin file path"
|
|
},
|
|
"working_directory": {
|
|
"type": "string",
|
|
"description": "Working directory"
|
|
},
|
|
"command": {
|
|
"type": "string",
|
|
"description": "Executed command"
|
|
},
|
|
"array_job_id": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Parent job ID for array jobs",
|
|
"nullable": true
|
|
},
|
|
"array_task_id": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Task ID within array job",
|
|
"nullable": true
|
|
}
|
|
}
|
|
},
|
|
"JobListResponse": {
|
|
"type": "object",
|
|
"properties": {
|
|
"jobs": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/JobResponse"
|
|
},
|
|
"description": "List of jobs"
|
|
},
|
|
"total": {
|
|
"type": "integer",
|
|
"description": "Total number of matching jobs"
|
|
},
|
|
"page": {
|
|
"type": "integer",
|
|
"description": "Current page number (starts at 1)"
|
|
},
|
|
"page_size": {
|
|
"type": "integer",
|
|
"description": "Items per page"
|
|
}
|
|
}
|
|
},
|
|
"NodeResponse": {
|
|
"type": "object",
|
|
"properties": {
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Node hostname"
|
|
},
|
|
"state": {
|
|
"type": "array",
|
|
"items": {
|
|
"type": "string"
|
|
},
|
|
"description": "Node state(s) (e.g. [\"IDLE\"], [\"ALLOCATED\",\"COMPLETING\"])"
|
|
},
|
|
"reason": {
|
|
"type": "string",
|
|
"description": "Reason for DOWN/DRAIN state"
|
|
},
|
|
"reason_set_by_user": {
|
|
"type": "string",
|
|
"description": "User who set the reason"
|
|
},
|
|
"cpus": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Total CPU cores"
|
|
},
|
|
"alloc_cpus": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Allocated CPU cores",
|
|
"nullable": true
|
|
},
|
|
"cores": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Physical cores",
|
|
"nullable": true
|
|
},
|
|
"sockets": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "CPU sockets",
|
|
"nullable": true
|
|
},
|
|
"threads": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Threads per core",
|
|
"nullable": true
|
|
},
|
|
"cpu_load": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "CPU load (kernel nice value * 100)",
|
|
"nullable": true
|
|
},
|
|
"real_memory": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Total physical memory (MiB)"
|
|
},
|
|
"alloc_memory": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Allocated memory (MiB)"
|
|
},
|
|
"free_mem": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Free memory (MiB)",
|
|
"nullable": true
|
|
},
|
|
"architecture": {
|
|
"type": "string",
|
|
"description": "System architecture (e.g. x86_64)"
|
|
},
|
|
"operating_system": {
|
|
"type": "string",
|
|
"description": "Operating system version"
|
|
},
|
|
"gres": {
|
|
"type": "string",
|
|
"description": "Available generic resources (e.g. \"gpu:4\")"
|
|
},
|
|
"gres_used": {
|
|
"type": "string",
|
|
"description": "Used generic resources (e.g. \"gpu:2\")"
|
|
},
|
|
"address": {
|
|
"type": "string",
|
|
"description": "Node address (IP)"
|
|
},
|
|
"hostname": {
|
|
"type": "string",
|
|
"description": "Node hostname (may differ from name)"
|
|
},
|
|
"weight": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Scheduling weight",
|
|
"nullable": true
|
|
},
|
|
"features": {
|
|
"type": "string",
|
|
"description": "Node feature tags (modifiable)"
|
|
},
|
|
"active_features": {
|
|
"type": "string",
|
|
"description": "Active feature tags (read-only)"
|
|
}
|
|
}
|
|
},
|
|
"PartitionResponse": {
|
|
"type": "object",
|
|
"properties": {
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Partition name"
|
|
},
|
|
"state": {
|
|
"type": "array",
|
|
"items": {
|
|
"type": "string"
|
|
},
|
|
"description": "Partition state(s) (e.g. [\"UP\"], [\"DOWN\",\"DRAIN\"])"
|
|
},
|
|
"default": {
|
|
"type": "boolean",
|
|
"description": "Whether this is the default partition"
|
|
},
|
|
"nodes": {
|
|
"type": "string",
|
|
"description": "Node range string"
|
|
},
|
|
"total_nodes": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Total nodes in partition"
|
|
},
|
|
"total_cpus": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Total CPUs in partition"
|
|
},
|
|
"max_cpus_per_node": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Max CPUs per node",
|
|
"nullable": true
|
|
},
|
|
"max_time": {
|
|
"type": "string",
|
|
"description": "Maximum time limit (\"UNLIMITED\" for unlimited)"
|
|
},
|
|
"max_nodes": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Max nodes per job",
|
|
"nullable": true
|
|
},
|
|
"min_nodes": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Min nodes per job",
|
|
"nullable": true
|
|
},
|
|
"default_time": {
|
|
"type": "string",
|
|
"description": "Default time limit"
|
|
},
|
|
"grace_time": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Grace time after preemption (seconds)",
|
|
"nullable": true
|
|
},
|
|
"priority": {
|
|
"type": "integer",
|
|
"format": "int32",
|
|
"description": "Job priority factor within partition",
|
|
"nullable": true
|
|
},
|
|
"qos_allowed": {
|
|
"type": "string",
|
|
"description": "Allowed QOS list"
|
|
},
|
|
"qos_deny": {
|
|
"type": "string",
|
|
"description": "Denied QOS list"
|
|
},
|
|
"qos_assigned": {
|
|
"type": "string",
|
|
"description": "Default assigned QOS"
|
|
},
|
|
"accounts_allowed": {
|
|
"type": "string",
|
|
"description": "Allowed accounts list"
|
|
},
|
|
"accounts_deny": {
|
|
"type": "string",
|
|
"description": "Denied accounts list"
|
|
}
|
|
}
|
|
},
|
|
"ParameterSchema": {
|
|
"type": "object",
|
|
"required": ["name", "type"],
|
|
"properties": {
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Parameter name"
|
|
},
|
|
"label": {
|
|
"type": "string",
|
|
"description": "Display label"
|
|
},
|
|
"type": {
|
|
"type": "string",
|
|
"enum": ["string", "integer", "enum", "file", "directory", "boolean"],
|
|
"description": "Parameter type"
|
|
},
|
|
"required": {
|
|
"type": "boolean",
|
|
"description": "Whether this parameter is required"
|
|
},
|
|
"default": {
|
|
"type": "string",
|
|
"description": "Default value"
|
|
},
|
|
"options": {
|
|
"type": "array",
|
|
"items": {
|
|
"type": "string"
|
|
},
|
|
"description": "Available options for enum type"
|
|
},
|
|
"description": {
|
|
"type": "string",
|
|
"description": "Parameter description"
|
|
}
|
|
}
|
|
},
|
|
"Application": {
|
|
"type": "object",
|
|
"properties": {
|
|
"id": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Application ID"
|
|
},
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Application name"
|
|
},
|
|
"description": {
|
|
"type": "string",
|
|
"description": "Application description"
|
|
},
|
|
"icon": {
|
|
"type": "string",
|
|
"description": "Application icon"
|
|
},
|
|
"category": {
|
|
"type": "string",
|
|
"description": "Application category"
|
|
},
|
|
"script_template": {
|
|
"type": "string",
|
|
"description": "Job script template"
|
|
},
|
|
"parameters": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/ParameterSchema"
|
|
},
|
|
"description": "Parameter definitions"
|
|
},
|
|
"scope": {
|
|
"type": "string",
|
|
"description": "Application scope",
|
|
"default": "system"
|
|
},
|
|
"created_by": {
|
|
"type": "integer",
|
|
"format": "int64",
|
|
"description": "Creator user ID",
|
|
"nullable": true
|
|
},
|
|
"created_at": {
|
|
"type": "string",
|
|
"format": "date-time",
|
|
"description": "Creation timestamp",
|
|
"nullable": true
|
|
},
|
|
"updated_at": {
|
|
"type": "string",
|
|
"format": "date-time",
|
|
"description": "Last update timestamp",
|
|
"nullable": true
|
|
}
|
|
}
|
|
},
|
|
"CreateApplicationRequest": {
|
|
"type": "object",
|
|
"required": ["name", "script_template"],
|
|
"properties": {
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Application name"
|
|
},
|
|
"description": {
|
|
"type": "string",
|
|
"description": "Application description"
|
|
},
|
|
"icon": {
|
|
"type": "string",
|
|
"description": "Application icon"
|
|
},
|
|
"category": {
|
|
"type": "string",
|
|
"description": "Application category"
|
|
},
|
|
"script_template": {
|
|
"type": "string",
|
|
"description": "Job script template"
|
|
},
|
|
"parameters": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/ParameterSchema"
|
|
},
|
|
"description": "Parameter definitions"
|
|
},
|
|
"scope": {
|
|
"type": "string",
|
|
"description": "Application scope",
|
|
"default": "system"
|
|
}
|
|
}
|
|
},
|
|
"UpdateApplicationRequest": {
|
|
"type": "object",
|
|
"properties": {
|
|
"name": {
|
|
"type": "string",
|
|
"description": "Application name"
|
|
},
|
|
"description": {
|
|
"type": "string",
|
|
"description": "Application description"
|
|
},
|
|
"icon": {
|
|
"type": "string",
|
|
"description": "Application icon"
|
|
},
|
|
"category": {
|
|
"type": "string",
|
|
"description": "Application category"
|
|
},
|
|
"script_template": {
|
|
"type": "string",
|
|
"description": "Job script template"
|
|
},
|
|
"parameters": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/ParameterSchema"
|
|
},
|
|
"description": "Parameter definitions"
|
|
},
|
|
"scope": {
|
|
"type": "string",
|
|
"description": "Application scope"
|
|
}
|
|
}
|
|
},
|
|
"ApplicationSubmitRequest": {
|
|
"type": "object",
|
|
"required": ["values"],
|
|
"properties": {
|
|
"values": {
|
|
"type": "object",
|
|
"additionalProperties": {
|
|
"type": "string"
|
|
},
|
|
"description": "Parameter name to value mapping"
|
|
}
|
|
}
|
|
},
|
|
"ApplicationListResponse": {
|
|
"type": "object",
|
|
"properties": {
|
|
"applications": {
|
|
"type": "array",
|
|
"items": {
|
|
"$ref": "#/components/schemas/Application"
|
|
},
|
|
"description": "List of applications"
|
|
},
|
|
"total": {
|
|
"type": "integer",
|
|
"description": "Total number of applications"
|
|
},
|
|
"page": {
|
|
"type": "integer",
|
|
"description": "Current page number"
|
|
},
|
|
"page_size": {
|
|
"type": "integer",
|
|
"description": "Items per page"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|