Get desktop application:
View/edit binary Protocol Buffers messages
Used in: , , ,
A backend that consists of a single endpoint.
A backend that comprises a load balanced service.
Describes queue configuration for a backend.
Describes a power-of-two-choices (P2C) load balancer configuration for a backend.
Used in:
The load estimation strategy used by this load balancer.
This load balancer uses peak EWMA (exponentially weighted moving average) load estimates.
A variant of peak EWMA that injects artificial penalties based on response codes, enabling 429-aware load balancing.
Parameters configuring peak EWMA load estimation.
Used in:
Initial latency value used when no latencies have been recorded for an endpoint.
The duration of the moving window over which latency is observed.
Used in:
Initial latency value used when no latencies have been recorded for an endpoint.
The duration of the moving window over which latency is observed.
The penalty duration to inject when a matching response is received.
The maximum duration to respect for Retry-After hints. Hints that exceed this value will be capped.
The EWMA decay window for the penalty.
A strategy for discovering endpoints for a service.
Used in:
Use the `Destination` service to discover endpoints for this service.
Used in:
Pool-level ejection protection. Prevents circuit breakers from ejecting all endpoints in a load-balancing pool. Additional ejection parameters may follow in future fields.
Used in:
Minimum number of endpoints that must remain unejected. When set to 0 (or absent), ejection protection is disabled.
Used in: , , ,
The minimum amount of time to wait before resuming an operation.
The maximum amount of time to wait before resuming an operation. Must be greater than or equal to min_backoff.
The ratio of the base timeout that may be randomly added to a backoff. Must be greater than or equal to 0.0.
If true, the proxy will use the Retry-After hint (if present on the response) as the initial backoff duration, clamped by min_backoff and max_backoff.
Configures failure accrual policies for circuit breaking. Setting a numeric policy field to zero disables that policy.
Used in: , ,
Ejection protection for the pool. When set, prevents circuit breakers from ejecting endpoints below the configured floor.
Used in:
Maximum consecutive failures before the circuit trips. Set to 0 to disable (an unset field has the same effect).
Must be set. Controls the ejection duration before probe requests are allowed after this policy trips the circuit.
Used in:
Success rate threshold in [0.0, 1.0]. The circuit trips when the EWMA success rate drops below this value. Set to 0.0 to disable. Defined as `double` (not `float`) because this value is compared against the proxy's f64 EWMA on every request.
EWMA decay window for success rate tracking.
Minimum requests before the success rate policy can trip (cold-start guard).
Maximum consecutive failures before the circuit trips. Set to 0 to disable (an unset field has the same effect).
Must be set. Controls the ejection duration before probe requests are allowed after this policy trips the circuit.
Used in:
If empty, the host value is ignored.
Must have at least one rule.
Used in:
Use the first available backend in the list.
Used in:
(message has no fields)
Used in:
Used in:
Used in: ,
Used in:
Must be set, even if there are no internal conditions.
Retryable gRPC status codes.
Used in:
Used in: ,
DEPRECATED: proxies ignore this. Retry timeouts are used instead of this.
Used in:
DEPRECATED: use `timeouts` instead. Servers should continue to set this value to the same value as `timeouts.response`.
If true, the proxy will allow headers to control retry and timeout behavior.
Used in:
Outbound-specific HTTP route configuration (based on the [Gateway API](https://gateway-api.sigs.k8s.io/v1alpha2/references/spec/#gateway.networking.k8s.io/v1alpha2.HTTPRoute)).
Used in: ,
If empty, the host value is ignored.
Must have at least one rule.
Used in:
Use the first available backend in the list.
Used in:
(message has no fields)
Used in:
Used in:
Used in: ,
Used in:
Must be set, even if there are no internal conditions.
Retryable conditions.
Used in:
Specifies the status code ranges that should trigger a retry.
Used in:
Used in: ,
DEPRECATED: proxies ignore this. Use Retry timeouts instead.
Used in:
DEPRECATED: use `timeouts` instead. Servers should continue to set this value to the same value as `timeouts.response`.
If true, the proxy will allow headers to control retry and timeout behavior.
Used in:
Used in:
Must have at least one rule.
Used in:
Use the first available backend in the list.
Used in:
(message has no fields)
Used in:
Used in:
Used in: ,
Used in: ,
Used in:
Used in:
Outbound policy for a given traffic spec.
Used as response type in: OutboundPolicies.Get, OutboundPolicies.Watch
Indicates the protocol to use for this target. This will be set to Opaque if the target has been marked as opaque and will be Discover otherwise.
Describes the resource for which outbound policy has been discovered.
Used in:
HTTP/1 policy configuration.
HTTP/2 policy configuration.
gRPC policy configuration.
TLS policy configuration.
Used in:
Protocol detection timeout.
HTTP/1 policy configuration.
HTTP/2 policy configuration.
Used in:
If empty, circuit breaking is not performed.
Used in: ,
If empty, circuit breaking is not performed.
Used in: ,
If empty, circuit breaking is not performed.
Used in: ,
Used in:
Used in:
The number of requests that may be held in a queue before backpressure is exerted.
A timeout that limits how long a backend may remain unready before any requests in its queue are failed.
Used in:
If empty, the SNI value is ignored.
Must have at least one rule.
Used in:
Use the first available backend in the list.
Used in:
(message has no fields)
Used in:
Used in:
Used in: ,
Used in: ,
Used in:
Used in:
Used as request type in: OutboundPolicies.Get, OutboundPolicies.Watch
Uniquely identifies the source proxy workload (e.g., pod name) to the control plane.
Describes a target address, as observed by the proxy.
Indicates the proxy is connecting to a specific IP:port.
Indicates the proxy is connecting to a named address (like an HTTP authority).