submit.proto

syntax = 'proto3';

package api;
option go_package = "github.com/armadaproject/armada/pkg/api";
option csharp_namespace = "ArmadaProject.Io.Api";

import "google/protobuf/empty.proto";
import "google/protobuf/timestamp.proto";
import "k8s.io/api/core/v1/generated.proto";
import "k8s.io/api/networking/v1/generated.proto";
import "google/api/annotations.proto";
import "pkg/api/health.proto";

message JobSubmitRequestItem {
    double priority = 1;
    string namespace = 3;
    string client_id = 8;
    map<string, string> labels = 4;
    map<string, string> annotations = 5;
    map<string, string> required_node_labels = 6 [deprecated = true]; // Use PodSpec.NodeSelector instead
    k8s.io.api.core.v1.PodSpec pod_spec = 2 [deprecated = true]; // Use PodSpecs instead
    repeated k8s.io.api.core.v1.PodSpec pod_specs = 7;
    repeated IngressConfig ingress = 9;
    repeated ServiceConfig services = 10;
    // Indicates which scheduler should manage this job.
    // If empty, the default scheduler is used.
    string scheduler = 11;
    // Ordinal 12 was previously used for queue_ttl_seconds
}

message IngressConfig {
    IngressType type = 1 [deprecated = true];
    repeated uint32 ports = 2;
    map<string, string> annotations = 3;
    bool tls_enabled = 4;
    string cert_name = 5;
    bool use_clusterIP = 6;
}

message ServiceConfig {
    ServiceType type = 1;
    repeated uint32 ports = 2;
}

// Ingress type is being kept here to maintain backwards compatibility for a while.
enum IngressType {
    Ingress = 0;
}

enum ServiceType {
    NodePort = 0;
    Headless = 1;
}

// swagger:model
message JobSubmitRequest {
    string queue = 1;
    string job_set_id = 2;
    repeated JobSubmitRequestItem job_request_items = 3;
}

// swagger:model
message JobPreemptRequest {
  string queue = 1;
  string job_set_id = 2;
  repeated string job_ids = 3;
  string reason = 4;
}

// swagger:model
message JobCancelRequest {
    string job_id = 1;
    string job_set_id = 2;
    string queue = 3;
    repeated string job_ids = 4;
    string reason = 5;
}

// swagger:model
message JobSetCancelRequest {
    string job_set_id = 1;
    string queue = 2;
    JobSetFilter filter = 3;
    string reason = 4;
}

// swagger:model
message JobSetFilter {
    repeated JobState states = 1;
}

// swagger:model
enum JobState {
    QUEUED = 0;
    PENDING = 1;
    RUNNING = 2;
    SUCCEEDED = 3;
    FAILED = 4;
    UNKNOWN = 5;
    SUBMITTED = 6;
    LEASED = 7;
    PREEMPTED = 8;
    CANCELLED = 9;
    REJECTED = 10;
}

message Job {
  string id = 1;
  string client_id = 13;
  string job_set_id = 2;
  string queue = 3;
  string namespace = 7;
  map<string, string> labels = 9;
  map<string, string> annotations = 10;
  map<string, string> required_node_labels = 11 [deprecated = true];
  string owner = 8;
  repeated string queue_ownership_user_groups = 15;
  bytes compressed_queue_ownership_user_groups = 19;
  double priority = 4;
  k8s.io.api.core.v1.PodSpec pod_spec = 5 [deprecated = true]; // Use PodSpecs instead
  repeated k8s.io.api.core.v1.PodSpec pod_specs = 12;
  // Resource requests and limits necessary for scheduling the main pod of this job.
  // The requests and limits herein are set to:
  //
  // max(
  //
  //	sum across all containers,
  //	max over all init containers,
  //
  // )
  //
  // This is because containers run in parallel, whereas initContainers run serially.
  // This field is populated automatically at submission.
  // Submitting a job with this field already populated results in an error.
  k8s.io.api.core.v1.ResourceRequirements scheduling_resource_requirements = 21;
  google.protobuf.Timestamp created = 6;
  // Services can be provided either as Armada-specific config objects or as proper k8s objects.
  // These options are exclusive, i.e., if either ingress or services is provided,
  // then neither of k8s_ingress or k8s_service can be provided, and vice versa.
  repeated IngressConfig ingress = 14;
  repeated ServiceConfig services = 16;
  repeated k8s.io.api.networking.v1.Ingress k8s_ingress = 17;
  repeated k8s.io.api.core.v1.Service k8s_service = 18;
  // Indicates which scheduler should manage this job.
  // If empty, the default scheduler is used.
  string scheduler = 20;
  // Queuing TTL for this job in seconds. If this job queues for more than this duration it will be cancelled. Zero indicates an infinite lifetime.
  int64 queue_ttl_seconds = 22;
}

// swagger:model
message JobReprioritizeRequest {
    repeated string job_ids = 1;
    string job_set_id = 2;
    string queue = 3;
    double new_priority = 4;
}

// swagger:model
message JobReprioritizeResponse {
    map<string, string> reprioritization_results = 1;
}

message JobSubmitResponseItem {
    string job_id = 1;
    string error = 2;
}

// swagger:model
message JobSubmitResponse {
    repeated JobSubmitResponseItem job_response_items = 1;
}

// swagger:model
message Queue {
    message Permissions {
        message Subject {
            string kind = 1;
            string name = 2;
        }

        repeated Subject subjects = 1;
        repeated string verbs = 2;
    }

    string name = 1;
    double priority_factor = 2;
    repeated string user_owners = 3;
    repeated string group_owners = 4;
    // These are ignored and should be removed
    map<string, double> resource_limits = 5 [deprecated = true];
    // Map from priority class name to resource limit overrides for this queue and priority class.
    // If provided for a priority class, global limits for that priority class do not apply to this queue.
    map<string, PriorityClassResourceLimits> resource_limits_by_priority_class_name = 7;
    repeated Permissions permissions = 6;
    // Determines whether scheduling is enabled for this queue.
    bool cordoned = 8;
    // A list of Kubernetes-like key-value labels, e.g. armadaproject.io/priority=critical
    repeated string labels_deprecated = 9 [deprecated=true];
    map<string,string> labels = 10;
}

message PriorityClassResourceLimits {
	// Limits resources assigned to jobs of this priority class.
	// Specifically, jobs of this priority class are only scheduled if doing so does not exceed this limit.
	map<string, double> maximum_resource_fraction = 1;
	// Per-pool override of maximum_resource_fraction.
	// If missing for a particular pool, maximum_resource_fraction is used instead for that pool.
	map<string, PriorityClassPoolResourceLimits> maximum_resource_fraction_by_pool = 2;
}

message PriorityClassPoolResourceLimits {
    map<string, double> maximum_resource_fraction = 1;
}

// swagger:model
message QueueList {
    repeated Queue queues = 1;
}

// swagger:model
message CancellationResult {
    repeated string cancelled_ids = 1;
}

//swagger:model
message QueueGetRequest {
    string name = 1;
}

//swagger:model
message QueueCordonRequest {
  string name = 1;
}

//swagger:model
message QueueUncordonRequest {
  string name = 1;
}

//swagger:model
message StreamingQueueGetRequest {
  uint32 num = 1;
}

//swagger:model
message QueueDeleteRequest {
    string name = 1;
}

message JobSetInfo {
    string name = 1;
    int32 queued_jobs = 2;
    int32 leased_jobs = 3;
}

message QueueUpdateResponse {
    Queue queue = 1;
    string error = 2;
}

message BatchQueueUpdateResponse {
    repeated QueueUpdateResponse failed_queues = 1;
}

message QueueCreateResponse {
    Queue queue = 1;
    string error = 2;
}

message BatchQueueCreateResponse {
    repeated QueueCreateResponse failed_queues = 1;
}

// Indicates the end of streams
message EndMarker{}

message StreamingQueueMessage{
  oneof event {
    Queue queue = 1;
    EndMarker end = 2;
  }
}

message QueuePreemptRequest {
  string name = 1;
  repeated string priorityClasses = 2;
}

message QueueCancelRequest {
  string name = 1;
  repeated string priorityClasses = 2;
  // In practice jobs are only cancellable from non-terminal states
  repeated JobState jobStates = 3;
}

service QueueService {
  rpc CreateQueue (Queue) returns (google.protobuf.Empty) {}
  rpc CreateQueues (QueueList) returns (BatchQueueCreateResponse) {}
  rpc UpdateQueue (Queue) returns (google.protobuf.Empty) {}
  rpc UpdateQueues (QueueList) returns (BatchQueueUpdateResponse){}
  rpc DeleteQueue (QueueDeleteRequest) returns (google.protobuf.Empty) {}
  rpc GetQueue (QueueGetRequest) returns (Queue) {}
  rpc GetQueues (StreamingQueueGetRequest) returns (stream StreamingQueueMessage) {}
  rpc CordonQueue (QueueCordonRequest) returns (google.protobuf.Empty) {}
  rpc UncordonQueue (QueueUncordonRequest) returns (google.protobuf.Empty) {}
  rpc PreemptOnQueue (QueuePreemptRequest) returns (google.protobuf.Empty) {}
  rpc CancelOnQueue (QueueCancelRequest) returns (google.protobuf.Empty) {}
}

service Submit {
    rpc SubmitJobs (JobSubmitRequest) returns (JobSubmitResponse) {
        option (google.api.http) = {
            post: "/v1/job/submit"
            body: "*"
        };
    }
    rpc CancelJobs (JobCancelRequest) returns (CancellationResult) {
        option (google.api.http) = {
            post: "/v1/job/cancel"
            body: "*"
        };
    }
    rpc CancelJobSet (JobSetCancelRequest) returns (google.protobuf.Empty) {
        option (google.api.http) = {
            post: "/v1/jobset/cancel"
            body: "*"
        };
    }
    rpc ReprioritizeJobs (JobReprioritizeRequest) returns (JobReprioritizeResponse) {
        option (google.api.http) = {
            post: "/v1/job/reprioritize"
            body: "*"
        };
    }
    rpc PreemptJobs (JobPreemptRequest) returns (google.protobuf.Empty) {
      option (google.api.http) = {
        post: "/v1/job/preempt"
        body: "*"
      };
    }
    rpc CreateQueue (Queue) returns (google.protobuf.Empty) {
        option (google.api.http) = {
            post: "/v1/queue"
            body: "*"
        };
    }
    rpc CreateQueues (QueueList) returns (BatchQueueCreateResponse) {
        option(google.api.http) = {
            post: "/v1/batched/create_queues"
            body: "*"
        };
    }
    rpc UpdateQueue (Queue) returns (google.protobuf.Empty) {
        option (google.api.http) = {
            put: "/v1/queue/{name}"
            body: "*"
        };
    }
    rpc UpdateQueues (QueueList) returns (BatchQueueUpdateResponse) {
        option (google.api.http) = {
            put: "/v1/batched/update_queues"
            body: "*"
        };
    }
    rpc DeleteQueue (QueueDeleteRequest) returns (google.protobuf.Empty) {
        option (google.api.http) = {
            delete: "/v1/queue/{name}"
        };
    }
    rpc GetQueue (QueueGetRequest) returns (Queue) {
        option (google.api.http) = {
            get: "/v1/queue/{name}"
        };
    }
    rpc GetQueues (StreamingQueueGetRequest) returns (stream StreamingQueueMessage) {
      option (google.api.http) = {
        get: "/v1/batched/queues"
      };
    }
    rpc Health(google.protobuf.Empty) returns (HealthCheckResponse);
}