package runtime

Get desktop application:
View/edit binary Protocol Buffers messages

ActivateComponentReply is a reply to an ActivateComponentRequest.

(message has no fields)

ActivateComponentRequest is a request from a weavelet to ensure that the provided component is running somewhere. An ActivateComponentRequest also implicitly signals that a weavelet is interested in receiving routing info for the component via UpdateRoutingInfoRequest.

• string component = 1

  component name

• bool routed = 2

  is the component routed?

AppConfig holds configuration for a Service Weaver application.

Used in: impl.BabysitterInfo, impl.SshConfig, multi.MultiConfig, Deployment, single.SingleConfig, status.Status

• string name = 1

  Application name (derived from binary if missing).

• string binary = 2

  File name of the binary that implements the app.

• repeated string args = 3

  Arguments to pass to the binary.

• repeated string env = 4

  Environment variables available to the binary. Every entry should be of the form "key=value". If a key appears more than once, only the last value of the key is used.

• repeated ComponentGroup colocate = 5

  Groups of components that must be co-located in the same OS process. This grouping allows the developer a level of control on where the Service Weaver runtime will place each component. For example, suppose that the application binary contains six components: A, B, C, D, E, and F. Furthermore, suppose that the developer specifies the following grouping in the config. colocate = [[A, B], [C, D, E]] In that case, no matter what algorithm the Service Weaver runtime uses to place the components, components A and B will always be grouped together, and components C, D, and E will always be grouped together. This means that a pair of replicas of the components can share in-memory state and communicate efficiently with each other (though not with other replicas). The placement algorithm the Service Weaver runtime uses varies across deployment environment. For example, a multiprocess deployer may choose to place each component into its own group, for maximum parallelism. A singleprocess deployer, on the other hand, may choose to group all of the components together. Regardless of which deployment environment is used, however, this configuration option guarantees that specified components will end up in the same group. Components are identified using their full package paths, e.g.: "github.com/my/project/package/ComponentName" with the exception of the component associated with the application main(), which is identified using the name "main".

• int64 rollout_nanos = 6

  A knob that lets the user specifies how many nanoseconds it expects for the rollout to take. This is used internally by Service Weaver to figure out the rollout strategy. If not specified, Service Weaver will pick a default value.

• map<string, string> sections = 7

  All config sections (includes [serviceweaver], [<deployer>], and [<component>] sections).

Assignment partitions a key space (e.g., the hash space [0, 2^64)) into a set of subregions, called slices, and assigns each slice to a set of replicas.

Used in: RoutingInfo

• repeated Assignment.Slice slices = 1

• uint64 version = 2

  The assignment for a component can change over time. Every one of these assignments is given a monotonically increasing version. Versions are unique for a given component, but not unique across components.

A slice of the key space and the replicas to which it is assigned.

Used in: Assignment

• uint64 start = 1

  Inclusive start key. Strictly greater than the `start` value for the previous entry in the current assignment. The first slice should always set start to 0.

• repeated string replicas = 2

  Replicas assigned to host this slice. Every address is of the form "<net>://<addr>" (e.g., "tcp://host:1234", "unix:///tmp/unix.sock").

A group of components.

Used in: AppConfig

• repeated string components = 1

Deployment holds internal information necessary for an application deployment. TODO(mwhittaker): Delete this proto.

• string id = 1

  Unique identifier for the deployment.

• optional AppConfig app = 2

  Application configuration.

• bool single_process = 10

  Run as a single process locally.

ExportListenerReply is a reply to an ExportListenerRequest.

• string proxy_address = 1

  The address of the proxy that forwards traffic to the listener, or empty if there is no proxy.

• string error = 2

  If non-empty, a non-retriable error. TODO(mwhittaker): This field exists as a hack to avoid a weavelet retrying this error. In the future, we should have a more robust way to avoid retrying certain errors.

ExportListenerRequest is a request from a weavelet to export the provided listener. Exporting a listener typically, but not always, involves running a proxy that forwards traffic to the provided address.

• string listener = 1

• string address = 2

GetHealthReply is a reply to a GetHealthRequest.

• HealthStatus status = 1

• repeated string healthy_components = 2

GetHealthRequest is a request from an envelope for a weavelet's health.

(message has no fields)

GetListenerAddressReply is a reply to a GetListenerAddressRequest.

• string address = 1

  address to listen on

GetListenerAddressRequest is a request from a weavelet for the address the weavelet should listen on for a particular listener. Imagine a weavelet calls GetListener with LocalAdress localhost:9000. Here are some examples of how different deployers may handle a GetListenerAddressRequest. - The singleprocess deployer may instruct the weavelet to listen directly on localhost:9000. - The multiprocess deployer may instruct the weavelet to listen on localhost:0. It will separately start a proxy on localhost:9000. - The SSH deployer may instruct the weavelet to listen on $HOSTNAME:0. It will separately start a proxy on localhost:9000.

• string name = 1

  listener name

GetLoadReply is a reply to a GetLoadRequest.

• optional LoadReport load = 1

GetLoadRequest is a request from an envelope for a weavelet's load report.

(message has no fields)

GetMetricsReply is a reply to a GetMetricsRequest. It only contains information about the metrics that have changed since the prior GetMetricsRequest.

• optional MetricUpdate update = 1

GetMetricsRequest is a request from an envelope for a weavelet's metrics. There can only be one outstanding GetMetricsRequest at a time.

(message has no fields)

GetProfileReply is a reply to a GetProfileRequest.

• bytes data = 1

  encoded profile data

GetProfileRequest is a request from an envelope for a weavelet to collect and return a profile. Some profile types only support a single profile request at a time and an error will be returned if another profile request is active.

• ProfileType profile_type = 1

  Type of the profile (e.g., heap, cpu).

• int64 cpu_duration_ns = 2

  Duration of CPU profiles, in nanoseconds.

GetSelfCertificateReply is a reply to a GetSelfCertificateRequest.

• bytes cert = 1

  PEM-encoded blocks containing (1) the certificate identifying the weavelet and (2) the corresponding private key. Used for enforcing mTLS security across components.

  certificate identifying the weavelet

• bytes key = 2

  private key corresponding to cert

GetSelfCertificateRequest is a request from a weavelet for its certificate and the corresponding private key.

(message has no fields)

HealthStatus specifies the health of a weavelet.

Used in: GetHealthReply

• UNKNOWN = 0

• HEALTHY = 1

• UNHEALTHY = 2

• TERMINATED = 3

InitWeaveletReply is the information provided by a weavelet to an envelope during the initial envelope-weavelet handshake.

• string dial_addr = 1

  Internal TCP address of the weavelet. This address will be passed to other weavelets as part of their RoutingInfo.

• optional SemVer version = 3

  The deployer API version of the weavelet.

InitWeaveletRequest holds the initialization info passed to the weavelet by the envelope.

• map<string, string> sections = 1

  See AppConfig.Sections.

LoadReport contains load information for all routed components hosted by a particular weavelet.

Used in: GetLoadReply

• map<string, LoadReport.ComponentLoad> loads = 1

  load for routed components

ComponentLoad includes load information for a component. The component's key space is divided into a number of slices, each of which is associated with a particular double-valued load. TODO(mwhittaker): In the short term, a weavelet will report one SliceLoad for every slice it is assigned. In the future, however, we may report load in more advanced ways.

Used in: LoadReport

• repeated SliceLoad load = 1

  TODO(mwhittaker): In full generality, load is a generic application-specific double-valued measurement (e.g., total number of requests, size of components in memory). For now to simplify things, we assume load is measured as requests per second.

• uint64 version = 2

  TODO(mwhittaker): For now, load is reported with respect to a given assignment. When the assignment changes, the load is reset. This approach works for rate-based load like requests per second, but not other non-rate based load like size of cached components.

SliceLoad includes the load of a slice, [start, end).

Used in: ComponentLoad

• uint64 start = 1

  inclusive

• uint64 end = 2

  exclusive

• double load = 3

• repeated SubsliceLoad splits = 4

• uint64 size = 5

  The "size" of the slice. Given a slice [a, b), it's tempting to think the slice has size b - a. This has a couple of problems. 1. Even if a slice is large, only a few keys in the slice may be used. For example, a weavelet with slice [0, 2^60) may only receive requests for keys in the range [0, 100). 2. a and b are not necessarily numeric. Imagine we are performing range sharding on strings and have the slice ["ccbae", "hhaefi"). What is the size of this slice? We instead measure the "size" of a slice as an estimate of the number of keys accessed in the slice. For example, if a slice [0, 1000) receives requests for keys 0, 1, 2, 3, and 10, then the size of the slice is 5.

SubsliceLoad divides the load of a slice into smaller subslices. For example, given a slice [0, 100) and load 9.0, we may have the following subslices: SubsliceLoad{Start: 0, Load: 3.0} // [0, 20) SubsliceLoad{Start: 20, Load: 3.0} // [20, 80) SubsliceLoad{Start: 80, Load: 3.0} // [80, 100) TODO(mwhittaker): In the future, we likely want to report hot keys as their own slice. Change SubsliceLoad to support that.

Used in: SliceLoad

• uint64 start = 1

• double load = 2

LogEntry is a log entry. Every log entry consists of a message (the thing the user logged) and a set of metadata describing the message.

Used in: LogEntryBatch

• string app = 1

  application name

• string version = 2

  deployment id

• string component = 3

  component name

• string node = 4

  weavelet id

• sfixed64 time_micros = 5

  time when logged (microseconds since epoch)

• string level = 6

  log level (e.g., debug, info, error)

• string file = 7

  source file

• int32 line = 8

  source line

• string msg = 9

  user provided message

• repeated string attrs = 10

  User-provided attributes and their associated values. The storage format is a name followed by its value, i.e.: name1, value1, name2, value2, name3, value3, ...

LogEntryBatch is a list of log entries.

• repeated LogEntry entries = 1

MetricDef is the definition of a new metric.

Used in: MetricUpdate

• uint64 id = 1

  metric's unique id

• string name = 2

  name of the metric

• MetricType typ = 3

  type of metric

• string help = 4

  metric's help message

• map<string, string> labels = 5

  metric labels

• repeated double bounds = 6

  histogram bucket bounds

MetricSnapshot is a snapshot of a metric. It is the union of a MetricDef and a MetricValue. TODO(mwhittaker): Move out of runtime.proto. It's not used by the pipe.

Used in: impl.BabysitterMetrics, status.Metrics

• uint64 id = 1

  metric's unique id

• string name = 2

  name of the metric

• MetricType typ = 3

  type of metric

• string help = 4

  metric's help message

• map<string, string> labels = 5

  metric labels

• repeated double bounds = 6

  histogram bucket bounds

• double value = 7

  value for counter and gauge, sum for histogram

• repeated uint64 counts = 8

  histogram counts

MetricType specifies a particular kind of metric.

Used in: MetricDef, MetricSnapshot

• INVALID = 0

• COUNTER = 1

• GAUGE = 2

• HISTOGRAM = 3

MetricUpdate contains metric updates.

Used in: GetMetricsReply

• repeated MetricDef defs = 1

  Metric definitions and updated values. All metrics are identified by their globally unique id. Two endpoints exchanging metric updates must ensure that a received metric value has been preceded by a metric definition with the same id, either in the current update or an earlier one.

  metric definitions

• repeated MetricValue values = 2

  metric values

MetricValue is the value associated with a metric.

Used in: MetricUpdate

• uint64 id = 1

  metric's unique id.

• double value = 2

  value for counter and gauge, sum for histogram

• repeated uint64 counts = 3

  histogram counts

ProfileType specifies a type of profile.

Used in: GetProfileRequest

• Unspecified = 0

• Heap = 1

• CPU = 2

RoutingInfo contains routing information for a component. A weavelet uses a RoutingInfo to know where to send a remote method call.

Used in: impl.GetRoutingInfoReply, UpdateRoutingInfoRequest

• string component = 1

• bool local = 2

  If true, perform method calls on the component locally. Replicas and assignment are ignored. Currently, local must always be true or always be false for a given component.

• repeated string replicas = 3

  The addresses of the weavelets hosting a replica of the component. Every address is of the form "<net>://<addr>" (e.g., "tcp://host:1234", "unix:///tmp/unix.sock"). For non-local components, a weavelet will execute method calls on one of these replicas.

• optional Assignment assignment = 4

  Routing assignment, if the component is routed.

SemVer represents a [semantic version][1] of the form "<major>.<minor>.<patch>". For example, the semver "0.12.3" has major=0, minor=12, and patch=3. [1]: https://semver.org/

Used in: InitWeaveletReply

• int64 major = 1

• int64 minor = 2

• int64 patch = 3

Span is a proto-based representation of an open-telemetry Span [1]. [1]: https://opentelemetry.io/docs/concepts/signals/traces/#spans TODO(spetrovic): Rename to TraceSpan.

Used in: TraceSpans

• string name = 1

  Span name

• bytes trace_id = 2

  Span trace id, 16 bytes long.

• bytes span_id = 3

  Unique span id, 8 bytes long.

• bytes parent_span_id = 4

  Parent span id, 8 bytes long

• Span.Kind kind = 5

• sfixed64 start_micros = 6

  Span start time (microseconds since epoch)

• sfixed64 end_micros = 7

  Span end time (microseconds since epoch)

• repeated Span.Attribute attributes = 8

• repeated Span.Link links = 9

  Links to other spans

• repeated Span.Event events = 10

  Events that occured in the span's lifetime

• optional Span.Status status = 11

  Span status

• optional Span.Scope scope = 18

• optional Span.Library library = 12

  Instrumentation library that created the span

• optional Span.Resource resource = 13

• int64 dropped_attribute_count = 14

  Number of span attributes dropped due to limits being reached.

• int64 dropped_link_count = 15

  Number of links dropped due to limits being reached.

• int64 dropped_event_count = 16

  Number of events dropped due to limits being reached.

• int64 child_span_count = 17

  Number of spans that consider this span a direct parent.

Span attributes.

Used in: Span, Event, Link, Resource

• string key = 1

• optional Attribute.Value value = 2

Used in: Attribute

• Value.Type type = 1

  Attribute type

• oneof value

  • uint64 num = 2

  • string str = 3

  • Value.NumberList nums = 4

  • Value.StringList strs = 5

Used in: Value

• repeated uint64 nums = 1

Used in: Value

• repeated string strs = 2

Type describes the type of the value.

Used in: Value

• INVALID = 0

  Used when no value is set

• BOOL = 1

• INT64 = 2

• FLOAT64 = 3

• STRING = 4

• BOOLLIST = 5

• INT64LIST = 6

• FLOAT64LIST = 7

• STRINGLIST = 8

Event represent an event that occurred in the span's lifetime.

Used in: Span

• string name = 1

  Even name

• sfixed64 time_micros = 2

  Time event was recorded (micros since epoch)

• repeated Attribute attributes = 3

  Event attributes

• int64 dropped_attribute_count = 4

  Number of attributes dropped due to limits being reached.

Span's role in the trace.

Used in: Span

• UNSPECIFIED = 0

• INTERNAL = 1

• SERVER = 2

• CLIENT = 3

• PRODUCER = 4

• CONSUMER = 5

Library stores information about the instrumentation library that created the span.

Used in: Span

• string name = 1

  Name of the instrumentation library, typically the package name of the library.

• string version = 2

  The version of the instrumentation library

• string schema_url = 3

  Schema url of the telemetry emitted by library

Link represents the link a span has to another span.

Used in: Span

• bytes trace_id = 1

  Span trace id

• bytes span_id = 2

  Unique span id

• repeated Attribute attributes = 3

  Span attributes

• int64 dropped_attribute_count = 4

  Number of span attributes dropped due to limits being reached

Resource describes an entity (e.g., a Service Weaver component) this span belongs to.

Used in: Span

• string schema_url = 1

  URL that describes the shape and a composition of data for this span.

• repeated Attribute attributes = 2

  Resource attributes

Scope stores information about the instrumentation scope that created the span.

Used in: Span

• string name = 1

  Name of the instrumentation library, typically the package name of the library.

• string version = 2

  The version of the instrumentation scope

• string schema_url = 3

  Schema url of the telemetry emitted by scope

Status represents the status for a span.

Used in: Span

• Status.Code code = 1

  Span status code

• string error = 2

  Error text; set only for an ERROR code

Used in: Status

• UNSET = 0

• ERROR = 1

• OK = 2

TraceSpans is a list of Span messages.

• repeated Span span = 1

UpdateComponentsReply is a reply to an UpdateComponentsRequest.

(message has no fields)

UpdateComponentsRequest is a request from an envelope to the weavelet to update the set of components it should be running. Currently, the set of components must only ever increase over time.

• repeated string components = 1

UpdateRoutingInfoReply is a reply to an UpdateRoutingInfoRequest.

(message has no fields)

UpdateRoutingInfoRequest is a request from an envelope to the weavelet to update its routing information for a particular component.

• optional RoutingInfo routing_info = 1

VerifyClientCertificateReply is a reply to a VerifyClientCertificateRequest. If the client certificate cannot be verified, an error should be returned rather than an empty reply.

• repeated string components = 1

  The set of components hosted by the weavelet that the client is allowed to invoke methods on.

VerifyClientCertificateRequest is a request from a weavelet to verify the identity of a client that is attempting to connect to the weavelet.

• repeated bytes cert_chain = 1

  Client's certificate chain. Each certificate in the chain is represented as a DER-encoded ASN.1 structure.

VerifyServerCertificateReply is a reply to a VerifyServerCertificateRequest.

(message has no fields)

VerifyServerCertificateRequest is a request from a weavelet to verify the identity of the server it is attempting to connect to.

• repeated bytes cert_chain = 1

  Server's certificate chain. Each certificate in the chain is represented as a DER-encoded ASN.1 structure.

• string target_component = 2

  Name of the component the server is meant to be hosting.

WeaveletArgs is the information provided by an envelope to a weavelet when the weavelet is created.

• string app = 1

  app name

• string deployment_id = 2

  globally unique deployment id

• string id = 3

  globally unique weavelet id

• bool run_main = 7

  run the main function?

• bool mtls = 8

  Should weavelets establish mTLS connections with each other?

• string internal_address = 10

  Address on which the weavelet's internal network listener should listen on (e.g., "localhost:12345", ":0"). If the address is empty, it defaults to ":0", like net.Listen. Note that for some deployers, the internal network listener can listen on an arbitrary port (don't set the port number). However, for deployers where listeners are prestarted (e.g., Kubernetes deployers), the port number should be propagated from the deployer.

• string control_socket = 13

  Unix domain socket path where weavelet should serve Control component method calls.

• repeated WeaveletArgs.Redirect redirects = 12

A redirect entry instructs the weavelet to direct calls made to component to be instead sent to the component named target at the specified address.

Used in: WeaveletArgs

• string component = 1

• string target = 2

• string address = 3