Menu
Platform
AI
App store purchases
Database
Flags
Jobs and cron
Localization
Monitoring
Notifications
Payments
Queues
Sandboxes
Webhooks
Getting Started
Authentication
KV Store
Deploy & Infrastructure
Reference
Create a distributed job
Creates a distributed job.
This method is not served on the public API.
api.sylphx.comdoes not route this call: its backend is not deployed behind the public API, or does not implement the call. This page documents the contract. It is kept out of the sidebar and of search engines.
Creates a distributed job.
- Path
POST https://api.sylphx.com/v1/orgs/acme/projects/shop/envs/production/distributed_jobs - Scope
workflows:write - Effect
write— a successful call changes state. - Collection distributed_jobs
- Query
distributed_job_id,validate_only - Operation answers with an
Operation; the result isDistributedJob
#Request
| Field | Type | What it is |
|---|---|---|
parent | string | The parent to create in. Required. |
distributed_job_id | string | The distributed job id, the final name segment: [a-z][a-z0-9-]{0,62}. |
distributed_job | DistributedJob | The distributed job to create; only caller-writable fields are read. Required. |
validate_only | bool | Validate and return the result without writing anything. |
#DistributedJob
| Field | Type | What it is |
|---|---|---|
name | string | orgs/{org}/projects/{project}/envs/{env}/distributed_jobs/{distributed_job}. |
meta | ResourceMeta | Resource metadata. |
spec | DistributedJobSpec | Desired state. Required. |
#ResourceMeta
| Field | Type | What it is |
|---|---|---|
labels | map<string, string> | Caller-writable, indexed labels (AIP-122 label rules). |
annotations | map<string, string> | Caller-writable, unindexed annotations. |
display_name | string | Caller-writable human-readable name. |
#DistributedJobSpec
| Field | Type | What it is |
|---|---|---|
worker | JobSpec | The worker container: its image, command, env and instance type. timeout is the longest one worker lives, so the pool turns over; max_retries and concurrency do not apply. Required. |
worker_class | WorkerClass | Price and preemption class; default scavenger. One of scavenger, standard. |
scale | WorkerScale | Elastic bounds of the worker pool. |
scratch_gib | int32 | Local scratch disk at /scratch per worker, GiB; 0 is none. |
shard_count | int64 | A run that brings no shards of its own gets this many index-only shards. |
shard_timeout | duration | The longest one lease may live without the shard ending; default 30m. |
shard_retries | int32 | Attempts a shard gets after a failure; a preemption is not one. Default 3. |
max_failed_shards | int64 | Failed shards the run tolerates before it fails; default 0. |
run_timeout | duration | The whole run's deadline; default 48h. |
checkpoint_uri | string | data://<bucket>/<prefix>/: checkpoints go under <run>/<shard>/. |
output_uri | string | data://<bucket>/<prefix>/: outputs go under <run>/<shard>/. |
#JobSpec
| Field | Type | What it is |
|---|---|---|
service | string | The Hosting Service whose env and Secret bindings every run takes, and whose live Release image it runs unless image or build names another, so a Job never runs code the environment does not serve. Required. |
image | string | A digest-pinned image from this org's Artifacts, in place of the service's image. Exclusive with build. |
command | string[] | The entrypoint; empty keeps the image's. |
args | string[] | Arguments; a run may replace them. |
env | map<string, string> | Environment overrides on top of the service's env and Secret bindings. |
instance_type | string | A Hosting instance type; default standard. |
timeout | duration | One attempt's deadline, 10s to 24h; default 1h. |
max_retries | int32 | Re-runs after a non-zero exit, 0 to 10; backoff 30s doubling. |
concurrency | JobConcurrency | What a start does while a run is active; default forbid. One of forbid, queue, allow. |
build | string | A build the project declares: its latest built image, in place of the service's; the image needs no running service of its own. Exclusive with image. |
#WorkerScale
| Field | Type | What it is |
|---|---|---|
min_workers | int32 | Workers kept while shards remain; default 1. |
max_workers | int32 | The most workers at once; the Cell and the org quota cap it too. Default 1. |
#Response
| Field | Type | What it is |
|---|---|---|
name | string | {parent-of-target}/operations/{operation}. |
target | string | The name of the Resource being changed. |
target_generation | int64 | The generation this Operation waits for. |
verb | OperationVerb | What the Operation does. One of create, update, delete, custom. |
custom_verb | string | The custom method verb when verb is CUSTOM, for example revoke. |
done | bool | Whether the Operation has finished. |
create_time | timestamp | When the Operation started. |
end_time | timestamp | When the Operation finished. |
response | any | The settled Resource (or Empty for a delete). One of the result group. |
error | ProblemDetails | The failure, as the one error body. One of the result group. |
progress | OperationProgress | Progress while not done. |
#ProblemDetails
| Field | Type | What it is |
|---|---|---|
type | string | https://sylphx.com/docs/errors/<code in kebab-case>. |
title | string | A short customer-safe summary of the problem class. |
status | int32 | The HTTP status, fixed by grpc_status. |
detail | string | A customer-safe explanation of this occurrence. Never contains SQL, stack traces, internal hostnames, or another tenant's data. |
instance | string | The request id (Sylphx-Request-Id). |
code | string | The stable UPPER_SNAKE code, one of ErrorCode without its prefix. |
grpc_status | string | One of the 16 canonical gRPC status names, for example ABORTED. |
retryable | bool | Whether the same request may be retried. |
effect | ErrorEffect | What the failed call committed. One of none, applied, unknown. |
retry_after_ms | int64 | How long to wait before retrying, when the server knows. |
details | any[] | Typed details such as PreconditionFailure or EntitlementDenial. |
#OperationProgress
| Field | Type | What it is |
|---|---|---|
percent | int32 | Percent complete, when the controller can tell. |
message | string | A customer-safe progress message. |
conditions | Condition[] | The target's conditions when this progress was read. |
#Condition
| Field | Type | What it is |
|---|---|---|
type | string | The condition type, for example Ready. |
status | ConditionStatus | Whether the condition holds. One of true, false, unknown. |
observed_generation | int64 | The generation this observation was made against. |
reason | string | A machine-readable UpperCamelCase reason. |
message | string | A customer-safe human-readable message. |
severity | Severity | How severe a FALSE condition is. One of info, warning, error. |
transition_time | timestamp | When status last changed. |
#Errors
UNAUTHENTICATED— No valid key or token was presented.PERMISSION_DENIED— The key lacks the method's permission.UNKNOWN_FIELD— The request has a field the schema does not know.INVALID_FIELD— A field failed validation.RESOURCE_ALREADY_EXISTS— A Resource with this name exists.IDEMPOTENCY_KEY_REUSED— An Idempotency-Key was reused with another body.IDEMPOTENCY_IN_PROGRESS— The first call with this Idempotency-Key is still running.PLAN_LIMIT_REACHED— The plan's limit is reached.
Every error arrives in the body Errors describes.
#Examples
curl -X POST "https://api.sylphx.com/v1/orgs/acme/projects/shop/envs/production/distributed_jobs" \
-H "Authorization: Bearer $SYLPHX_API_KEY" \
-H "Content-Type: application/json" \
-d '{"spec":{"worker":{"service":"…"}}}'