Post

/pipeline-revisions

Body parameters

object

Responses

201

The created Pipeline Revision

Expand All

object

name

string

A name for the resource.

Examplenaomis-resource

description

string

A description of the resource.

ExampleA good and detailed description that makes everyone understand.

tags

Array of string

Custom tags for organizing and categorizing resources in the GUI.

Max items32

Example[ "tagA", "tagB", "tagC" ]

string

id

integer (int64)

A globally unique identifier (ID) for the resource.

guid

string (uuid)

A globally unique identifier (GUID) for the resource.

Examplea7f4aa24-a00d-4dc5-a113-2031eb8c2c79

tenant_guid

string (uuid)

A globally unique identifier (GUID) for the resource.

Examplea7f4aa24-a00d-4dc5-a113-2031eb8c2c79

owner

object (UserID)

Represents the owner of the resource.

id

string

The unique identifier for the owner.

id_type

string

The type of identifier for the owner.

Valid values[ "uid", "vid", "sid", "access_key", "sid_str", "login_name", "username" ]

name

string

The presentable name of the owner

created_at

string (date-time)

Timestamp when the resource was created.

Example2024-12-25T15:30:00Z

updated_at

string (date-time)

Timestamp when the resource was last updated.

Example2024-12-28T18:00:00Z

vrn

string

A logical path for resource grouping and policy assignment. The structure of the logical path is: vast:dataengine:: where resource type is one of: triggers / functions / pipelines / topics / kubernetes-clusters / container-registries / mtls-authentication-credentials.

Examplevast:dataengine:functions:foo

revision_number

integer

The latest revision number

Minimum0

Example69

pipeline_guid

string (uuid)

A globally unique identifier (GUID) for the resource.

Examplea7f4aa24-a00d-4dc5-a113-2031eb8c2c79

manifest

object (Manifest)

Example{ "triggers": [ { "name": "trigger1", "vrn": "vast:dataengine:triggers:trigger1" } ], "function_deployments": [ { "name": "function1", "vrn": "vast:dataengine:functions:function1", "revision": 1, "config": { "environment_variables": [ { "name": "TARGET", "value": "function2" } ], "log_level": "INFO", "event_batching": { "batch_max_size": 100, "batch_timeout_ms": 5000 }, "timeout": 60 }, "resources": { "min_concurrency": 1, "max_concurrency": 10, "min_cpu": "100m", "max_cpu": "500m", "min_memory": "128Mi", "max_memory": "256Mi", "timeout": 60 } }, { "name": "function2", "vrn": "vast:functions:function2" } ], "links": [ { "source": [ "trigger1" ], "destination": [ "function1" ], "topic": "vast:dataengine:topics:topic1", "config": { "retries": 3, "events_order": "ordered" }, "conditional_trigger_labels": [ { "triggerLabelName": "logginglevel", "triggerLabelValue": "debug" }, { "triggerLabelName": "errorseverity", "triggerLabelValue": "high" } ] }, { "source": [ "function1" ], "destination": [ "function2" ], "topic": "vast:dataengine:topics:topic2" } ] }

config

object

environment_variables

Array of object (EnvironmentVariable)

object

name

string

Min length1

value

string

Min length1

secrets

Array of string

string

triggers

Array of object

object

name

string

vrn

string

A logical path for resource grouping and policy assignment. The structure of the logical path is: vast:dataengine:: where resource type is one of: triggers / functions / pipelines / topics / kubernetes-clusters / container-registries / mtls-authentication-credentials.

Examplevast:dataengine:functions:foo

function_deployments

Array of object

object

name

string

Max length40

function_vrn

string

A logical path for resource grouping and policy assignment. The structure of the logical path is: vast:dataengine:: where resource type is one of: triggers / functions / pipelines / topics / kubernetes-clusters / container-registries / mtls-authentication-credentials.

Examplevast:dataengine:functions:foo

revision

integer | null

config

object

Example{ "environment_variables": [ { "name": "TARGET", "value": "function2" } ] }

environment_variables

Array of object (EnvironmentVariable)

object

name

string

Min length1

value

string

Min length1

secrets

Array of string

string

event_batching

object (EventBatchingConfig)

batch_max_size

integer

The maximum number of events to batch together.

Minimum1

Maximum512

Example100

batch_timeout_ms

integer

The timeout in milliseconds for the batch. The batch will be invoked after this timeout, even if the batch is not full.

Minimum0

Maximum60000

Example5000

log_level

string

Valid values[ "NOTSET", "DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL" ]

ExampleINFO

resources

object (FunctionDeploymentResources)

Example{ "min_concurrency": 0, "max_concurrency": 10, "min_cpu": "100m", "max_cpu": "500m", "min_memory": "128Mi", "max_memory": "256Mi" }

min_concurrency

integer

Default1

Example1

max_concurrency

integer

10

Default10

autoscaling_rps_factor

integer | null Deprecated

Deprecated: This field is no longer used and will be removed in a future version. It is ignored if provided.

Example2

autoscaling_concurrency_overload_factor

string | null Deprecated

Deprecated: Use 'autoscaling' field instead with type='kpa_concurrency'.

Autoscaling concurrency overload factor (> 0.0). Determines when to scale based on concurrency overload beyond current processing capacity. When concurrency exceeds (capacity * (1 + overload_factor)), new function instances are created.

For backward compatibility, if this field is set and 'autoscaling' is not provided, concurrency metric autoscaling will be used with target_utilization_percentage calculated from this value.

Default: "0.05" (5% overload tolerance)

Pattern^(0|[1-9]\d*)(\.\d+)?$

Example0.05

autoscaling

object (AutoscalingConfig)

Autoscaling configuration for the function deployment. Defines how the function scales based on incoming requests or resource utilization.

Three autoscaling metrics are supported:

concurrency: Scale based on concurrent requests per instance (default, supports scale-to-zero)
rps: Scale based on requests per second per instance (supports scale-to-zero)
resource: Scale based on CPU or memory utilization (does not support scale-to-zero)

Only one of 'concurrency', 'rps', or 'resource' should be provided, matching the 'metric' field.

Example{ "metric": "concurrency", "concurrency": { "target": 20, "target_utilization_percentage": 70 } }

metric

string

The metric used for autoscaling decisions. Determines which configuration object should be provided:

concurrency: Provide 'concurrency' configuration object
rps: Provide 'rps' configuration object
resource: Provide 'resource' configuration object

Valid values[ "concurrency", "rps", "resource" ]

Exampleconcurrency

concurrency

object

Autoscaling configuration based on concurrent requests metric. Extends RequestAutoscalingConfig with concurrency-specific options.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

hard_limit

integer

Hard limit for concurrent requests per instance. Surplus requests will be buffered until capacity is available. Set to 0 for unlimited (default).

Warning: Using a hard limit may negatively impact throughput and latency. Only recommended if there's a clear use case for limiting concurrency.

Minimum0

Example0

rps

object (RequestAutoscalingConfig)

Common autoscaling configuration for request-based metrics (concurrency and RPS). This configuration supports scale-to-zero when there is no traffic.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

resource

object (ResourceAutoscalingConfig)

Autoscaling configuration based on resource utilization (CPU or memory).

Note: Resource-based autoscaling does not support scale-to-zero. Use concurrency or rps metrics if scale-to-zero is required.

Example{ "resource_autoscaling_type": "cpu", "target": 80 }

resource_autoscaling_type

string

The resource type to use for autoscaling:

cpu: Scale based on CPU utilization percentage
memory: Scale based on memory utilization percentage

Valid values[ "cpu", "memory" ]

Examplecpu

target

integer

Target value for the selected resource. For CPU, this is the target utilization percentage (e.g., 80 means 80% of CPU request). Note: The Knative documentation at https://knative.dev/docs/serving/autoscaling/autoscaling-metrics/ incorrectly states this is in millicores, but testing and Knative source code confirm it's a percentage. For memory, this is in Mi (e.g., 200 means 200Mi). The HPA will scale when actual usage exceeds this target.

Minimum1

Example80

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). This controls the HPA stabilization window for scale-down decisions. Larger values result in more smoothing and slower scale-down. Default: not set (uses Kubernetes HPA default of 300s)

Minimum6

Maximum3600

Example60

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

hard_limit

integer

Hard limit for concurrent requests per instance. Surplus requests will be buffered until capacity is available. Set to 0 for unlimited (default).

Warning: Using a hard limit may negatively impact throughput and latency. Only recommended if there's a clear use case for limiting concurrency.

Minimum0

Example0

concurrency

object

Autoscaling configuration based on concurrent requests metric. Extends RequestAutoscalingConfig with concurrency-specific options.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

hard_limit

integer

Hard limit for concurrent requests per instance. Surplus requests will be buffered until capacity is available. Set to 0 for unlimited (default).

Warning: Using a hard limit may negatively impact throughput and latency. Only recommended if there's a clear use case for limiting concurrency.

Minimum0

Example0

rps

object (RequestAutoscalingConfig)

Common autoscaling configuration for request-based metrics (concurrency and RPS). This configuration supports scale-to-zero when there is no traffic.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

resource

object (ResourceAutoscalingConfig)

Autoscaling configuration based on resource utilization (CPU or memory).

Note: Resource-based autoscaling does not support scale-to-zero. Use concurrency or rps metrics if scale-to-zero is required.

Example{ "resource_autoscaling_type": "cpu", "target": 80 }

resource_autoscaling_type

string

The resource type to use for autoscaling:

cpu: Scale based on CPU utilization percentage
memory: Scale based on memory utilization percentage

Valid values[ "cpu", "memory" ]

Examplecpu

target

integer

Target value for the selected resource. For CPU, this is the target utilization percentage (e.g., 80 means 80% of CPU request). Note: The Knative documentation at https://knative.dev/docs/serving/autoscaling/autoscaling-metrics/ incorrectly states this is in millicores, but testing and Knative source code confirm it's a percentage. For memory, this is in Mi (e.g., 200 means 200Mi). The HPA will scale when actual usage exceeds this target.

Minimum1

Example80

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). This controls the HPA stabilization window for scale-down decisions. Larger values result in more smoothing and slower scale-down. Default: not set (uses Kubernetes HPA default of 300s)

Minimum6

Maximum3600

Example60

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

concurrency

object

Autoscaling configuration based on concurrent requests metric. Extends RequestAutoscalingConfig with concurrency-specific options.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

hard_limit

integer

Hard limit for concurrent requests per instance. Surplus requests will be buffered until capacity is available. Set to 0 for unlimited (default).

Warning: Using a hard limit may negatively impact throughput and latency. Only recommended if there's a clear use case for limiting concurrency.

Minimum0

Example0

rps

object (RequestAutoscalingConfig)

Common autoscaling configuration for request-based metrics (concurrency and RPS). This configuration supports scale-to-zero when there is no traffic.

Example{ "target": 20, "target_utilization_percentage": 70, "stable_window_seconds": 60, "burst_threshold_percentage": 200 }

target

integer

Target value per instance. The autoscaler will try to maintain this level per instance.

For concurrency metric: target number of concurrent requests (default: 20, which matches the default number of worker threads in non-asyncio Python functions)
For RPS metric: target requests per second

Minimum1

Default20

Example20

target_utilization_percentage

integer

Target utilization percentage (1-100). Specifies what percentage of the target should actually be targeted by the autoscaler. This causes the autoscaler to scale up before the target is reached.

For example, with target=10 and target_utilization_percentage=70, the autoscaler will create a new instance when average metric value reaches 7.

Minimum1

Maximum100

Default70

Example70

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). Larger values result in more smoothing.

When scaling to zero, the last instance is only removed after no traffic for the entire duration of this window.

Minimum6

Maximum3600

Default60

Example60

burst_window_percentage

integer

Burst detection window as a percentage of stable window (1-100). Defines the shorter time window used to detect traffic bursts. Lower values make burst detection more sensitive.

Minimum1

Maximum100

Default10

Example10

burst_threshold_percentage

integer

Burst mode threshold as a percentage of current capacity (110-1000). Burst mode is triggered when traffic exceeds this percentage of what current instances can handle.

For example, 200 means burst mode starts when traffic is 2x current capacity.

Minimum110

Maximum1000

Default200

Example200

scale_down_delay_seconds

integer

Delay in seconds before scaling down after load decreases. Helps prevent thrashing during fluctuating traffic.

Minimum0

Default300

Example300

scale_to_zero_delay_seconds

integer

Minimum time in seconds to retain the last instance before scaling to zero. The instance is kept for at least this duration after the autoscaler decides to scale to zero.

Minimum0

Default600

Example600

resource

object (ResourceAutoscalingConfig)

Autoscaling configuration based on resource utilization (CPU or memory).

Note: Resource-based autoscaling does not support scale-to-zero. Use concurrency or rps metrics if scale-to-zero is required.

Example{ "resource_autoscaling_type": "cpu", "target": 80 }

resource_autoscaling_type

string

The resource type to use for autoscaling:

cpu: Scale based on CPU utilization percentage
memory: Scale based on memory utilization percentage

Valid values[ "cpu", "memory" ]

Examplecpu

target

integer

Target value for the selected resource. For CPU, this is the target utilization percentage (e.g., 80 means 80% of CPU request). Note: The Knative documentation at https://knative.dev/docs/serving/autoscaling/autoscaling-metrics/ incorrectly states this is in millicores, but testing and Knative source code confirm it's a percentage. For memory, this is in Mi (e.g., 200 means 200Mi). The HPA will scale when actual usage exceeds this target.

Minimum1

Example80

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). This controls the HPA stabilization window for scale-down decisions. Larger values result in more smoothing and slower scale-down. Default: not set (uses Kubernetes HPA default of 300s)

Minimum6

Maximum3600

Example60

resource_autoscaling_type

string

The resource type to use for autoscaling:

cpu: Scale based on CPU utilization percentage
memory: Scale based on memory utilization percentage

Valid values[ "cpu", "memory" ]

Examplecpu

target

integer

Target value for the selected resource. For CPU, this is the target utilization percentage (e.g., 80 means 80% of CPU request). Note: The Knative documentation at https://knative.dev/docs/serving/autoscaling/autoscaling-metrics/ incorrectly states this is in millicores, but testing and Knative source code confirm it's a percentage. For memory, this is in Mi (e.g., 200 means 200Mi). The HPA will scale when actual usage exceeds this target.

Minimum1

Example80

stable_window_seconds

integer

Time window in seconds for stable autoscaling decisions (6 to 3600). This controls the HPA stabilization window for scale-down decisions. Larger values result in more smoothing and slower scale-down. Default: not set (uses Kubernetes HPA default of 300s)

Minimum6

Maximum3600

Example60

num_workers

integer

Number of worker threads per function instance.
When event batching is enabled, the effective concurrency per instance is num_workers * batch_max_size, since each worker processes one batch at a time. Default: 20 (matches the default Python runtime worker threads) Maximum: 100

Minimum1

Maximum100

Default20

Example20

max_concurrent_requests_limit

integer

Maximum number of concurrent requests the function can handle per instance. This represents how many requests can be processed simultaneously (in-flight) by a single function instance. Maps to Knative spec.containerConcurrency.

If omitted, the value is computed at deploy time as max(num_workers, batch_max_size) (defaults: max(20, 1) = 20). This ensures the queue-proxy enforces backpressure per pod so the autoscaler can scale out instead of all traffic piling on a single pod.

Set explicitly to 0 for unlimited per-pod concurrency (queue-proxy does not throttle). Use with caution: with 0 the autoscaler cannot observe queue pressure and may keep traffic on a single instance.

Setting this value below the actual function capacity will hurt performance. Set ONLY if you know an upper bound for the capacity; this applies backpressure on a loaded function instance and waits for autoscale instead of queueing requests, which can tune performance and avoid timed-out requests in the queue and retries.

Minimum0

Example100

min_cpu

string

Default"100m"

Example100m

max_cpu

string

Default"500m"

Example500m

min_memory

string

Default"128Mi"

Example128Mi

max_memory

string

Default"256Mi"

Example256Mi

storage

string

Example512Mi

timeout

integer

Timeout in seconds for the function deployment.

Example60

links

Array of object

Links which define the calling graph of functions in a pipeline. | Each of the sources at the source array triggers all function deployments in the destination array | if all the labels in the conditional_trigger_labels array are set to the specified values. | The label values are set by the source function handler using the context.set_trigger_labels() method.

object

source

Array of string

string

destination

Array of string

string

topic

string

A logical path for resource grouping and policy assignment. The structure of the logical path is: vast:dataengine:: where resource type is one of: triggers / functions / pipelines / topics / kubernetes-clusters / container-registries / mtls-authentication-credentials.

Examplevast:dataengine:functions:foo

config

object (LinkConfig)

retries

integer

Number of retries for the trigger.

Example3

events_order

string

Ordering of events for the trigger.

Valid values[ "ordered", "unordered" ]

conditional_trigger_labels

Array of object

Conditional trigger labels for the link. These are used to conditionally trigger the link. | All labels must be set to the specified values to trigger the link (i.e. AND operation between labels) | Labels names must be up to 15 lowercase characters or digits, must start with a letter

Example[ { "triggerLabelName": "logginglevel", "triggerLabelValue": "debug" }, { "triggerLabelName": "errorseverity", "triggerLabelValue": "high" } ]

object

triggerLabelName

string

Pattern^[a-z][a-z0-9]{0,14}$

triggerLabelValue

string

Pattern^[a-zA-Z0-9]{1,64}$

deployed

boolean

If the pipeline revision was deployed

Defaultfalse

Create a new Pipeline Revision