types.Execution
DeploymentID tracks which deployment this execution belongs to
EvalID is the ID of the evaluation that generated this execution
ID is the unique identifier of the execution (UUID)
JobID is the ID of the job that this execution is for
job_spec object
JobSpec is the parent job of the task being allocated. This is copied at execution time to avoid issues if the job definition is updated. TODO: Consider removing this field to reduce storage size, and populate it on-demand from the job store.
config object
Config contains type-specific configuration for the workload. The structure depends on the job type (e.g., pipeline config, query parameters).
Config contains type-specific configuration for the workload. The structure depends on the job type (e.g., pipeline config, query parameters).
deployment object
Deployment defines how to deploy the job
Auto-promote canary deployments
Canary-specific settings
Percentage of canary nodes
health_check object
HealthCheck defines health check configuration (required for rolling/canary, ignored for immediate)
Deadline is the maximum time to wait for an execution to become healthy (required) Example: "5m", "30s"
MaxErrorRate is the maximum error rate allowed during health checks (optional, default: 0.10) Example: 0.10 = 10% error rate
MinHealthyTime is the minimum continuous healthy duration required (optional, default: "30s") Example: "30s", "1m"
MaxFailedNodes is the maximum number of failed nodes before stopping (optional, default: 10)
MaxFailedNodesPercent is the maximum percentage of failed nodes before stopping (optional, default: 10.0)
MaxParallel is the maximum number of nodes to update in parallel (0 = all at once)
NoAutoRollback disables automatic rollback on deployment failure (default: false = auto-rollback enabled)
Strategy: immediate|rolling|canary
Possible values: [immediate, rolling, canary]
Description is an optional human-readable description of the job.
labels object
Labels is used to associate arbitrary labels with this job. Labels can be used for filtering and selection.
meta object
Meta is used to associate arbitrary metadata with this job. Keys with the prefix "expanso.io/" are reserved for system use.
Name is the logical name of the job used to refer to it. Submitting a job with the same name as an existing job will result in an update to the existing job.
Namespace is the namespace this job is running in.
Priority defines the scheduling priority of this job. Higher values indicate higher priority.
selector object
Selector defines which nodes to run the job on
MatchExpressions selects nodes using label selector expression strings. Each expression is evaluated independently and all must match (AND logic). Supported syntax:
- Equality: "key=value" or "key==value"
- Inequality: "key!=value"
- Set inclusion: "key in (value1,value2,...)"
- Set exclusion: "key notin (value1,value2,...)"
- Existence: "key"
- Non-existence: "!key" Examples:
- "region=us-east"
- "tier in (premium,standard)"
- "environment!=prod"
- "gpu"
- "!debug"
MatchIDs selects specific nodes by their IDs. If specified, the job will only run on nodes whose ID is in this list.
match_labels object
MatchLabels selects nodes with labels that exactly match all specified key-value pairs. All labels must match (AND logic). Example: {"region": "us-east", "tier": "compute"}
timeouts object
Timeouts defines timeout configurations for the job
ExecutionTimeout is the maximum amount of time a task is allowed to run in seconds. Zero means no timeout, such as for a daemon task.
QueueTimeout is the maximum amount of time a task is allowed to wait in the orchestrator queue in seconds before being scheduled. Zero means no timeout.
TotalTimeout is the maximum amount of time a task is allowed to complete in seconds. This includes the time spent in the queue, the time spent executing and the time spent retrying. Zero means no timeout.
Type specifies what kind of workload this job runs (e.g. "pipeline", "query", "update", "config"). The scheduling behavior is derived from this type.
JobType is the type of job this execution is for
JobVersion is the version of the job when this execution was created
Namespace is the namespace the execution is created in
NodeID is the node this execution is placed on
status object
Status contains system-managed fields for the execution
compute_state object
ComputeState is the actual state of the execution on the node
details object
Details is a map of additional details about the state.
Message is a human readable message describing the state.
StateType is the current state of the object.
Possible values: [pending, starting, running, degraded, complete, failed, lost, stopping, stopped]
CreatedAt is when the execution was created
desired_state object
DesiredState is what state the execution should be in
details object
Details is a map of additional details about the state.
Message is a human readable message describing the state.
StateType is the current state of the object.
Possible values: [pending, running, stopped]
details object
Details contains structured metadata about the current execution state (e.g., error codes, hints, component info from lib/errors)
NextExecutionID is used for tracking and triggering the next execution, such as during an update
PreviousExecutionID is used for tracking the previous execution in a sequence
Revision is incremented on any change to the execution (state changes, status updates, etc.)
UpdatedAt is when the execution was last updated
{
"id": "exec-789xyz456abc",
"job_id": "job-abc123xyz",
"job_type": "pipeline",
"job_version": 2,
"namespace": "production",
"node_id": "node-edge-us-west-01",
"status": {
"compute_state": {
"message": "Pipeline processing logs",
"state_type": "running"
},
"created_at": "2025-01-15T10:35:00Z",
"desired_state": {
"message": "",
"state_type": "run"
},
"revision": 5,
"updated_at": "2025-01-15T10:36:15Z"
}
}