Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func FleetStreams ¶
func FleetStreams() []string
FleetStreams are the streams the fleet owns in the store. Keeping them named here means a pipeline cannot quietly write into a stream nobody declared.
Types ¶
type Collector ¶
type Collector struct {
Implementation string `json:"implementation" yaml:"implementation"`
Repository string `json:"repository" yaml:"repository"`
Version string `json:"version" yaml:"version"`
Architecture string `json:"architecture" yaml:"architecture"`
Archive Asset `json:"archive" yaml:"archive"`
Checksum Asset `json:"checksum" yaml:"checksum"`
SBOM Asset `json:"sbom" yaml:"sbom"`
SigstoreBundle Asset `json:"sigstore_bundle" yaml:"sigstore_bundle"`
BinaryPath string `json:"binary_path" yaml:"binary_path"`
DeploymentStage string `json:"deployment_stage" yaml:"deployment_stage"`
}
Collector pins the OpenTelemetry Collector that runs beside the fleet control plane. Upstream publishes a SHA-256, a CycloneDX SBOM and a Sigstore bundle for every asset, so all three are pinned rather than only the archive.
type Durability ¶
type Durability struct {
RPO string `json:"rpo" yaml:"rpo"`
BackendOutageMinutes int `json:"backend_outage_minutes" yaml:"backend_outage_minutes"`
RecoveryRTOMinutes int `json:"recovery_rto_minutes" yaml:"recovery_rto_minutes"`
RetryMaxElapsedTime string `json:"retry_max_elapsed_time" yaml:"retry_max_elapsed_time"`
QueueSizeBatchesPerSignal int `json:"queue_size_batches_per_signal" yaml:"queue_size_batches_per_signal"`
BlockOnOverflow bool `json:"block_on_overflow" yaml:"block_on_overflow"`
PersistentStorage string `json:"persistent_storage" yaml:"persistent_storage"`
AcceptedLossOutside string `json:"accepted_loss_outside_envelope" yaml:"accepted_loss_outside_envelope"`
RequiredSelfMetrics []string `json:"required_self_metrics" yaml:"required_self_metrics"`
Note string `json:"note" yaml:"note"`
}
type Manifest ¶
type Manifest struct {
SchemaVersion int `json:"schema_version" yaml:"schema_version"`
Collector Collector `json:"collector" yaml:"collector"`
Store Store `json:"store" yaml:"store"`
Transport Transport `json:"transport" yaml:"transport"`
}
Manifest pins every release input of the telemetry plane. It contains no credentials and never treats a mutable release alias as an artifact, so a `latest` tag or an unversioned download URL is rejected rather than trusted.
The plane has exactly two components. The collector runs on the CI host, which deliberately has no container runtime in its control plane, so it is pinned as a release tarball the same way the runner and sccache are. The store runs on a separate host and is pinned by image digest, because that is the only immutable identity a registry offers.
func (Manifest) CanonicalJSON ¶
func (Manifest) Fingerprint ¶
type Store ¶
type Store struct {
Implementation string `json:"implementation" yaml:"implementation"`
Version string `json:"version" yaml:"version"`
Archive Asset `json:"archive" yaml:"archive"`
BinaryPath string `json:"binary_path" yaml:"binary_path"`
Host string `json:"host" yaml:"host"`
Organization string `json:"organization" yaml:"organization"`
ObjectStore string `json:"object_store" yaml:"object_store"`
Bucket string `json:"bucket" yaml:"bucket"`
Streams []string `json:"streams" yaml:"streams"`
RetentionDays int `json:"retention_days" yaml:"retention_days"`
// StreamRetentionDays overrides RetentionDays for one stream. The two
// signals age differently: a log line is worth keeping while someone is
// still debugging the week it came from, and a metric point is the only
// evidence of what the fleet's capacity did over a quarter. Storing both
// for the same 90 days meant the class sizing done from seven days of
// history had no longer window to check itself against, while nine
// million log records were kept long past anyone reading them.
StreamRetentionDays map[string]int `json:"stream_retention_days,omitempty" yaml:"stream_retention_days,omitempty"`
DeploymentStage string `json:"deployment_stage" yaml:"deployment_stage"`
}
Store pins the OpenObserve instance that receives the fleet's telemetry.
It is pinned as a release tarball rather than a container image because it now runs on a fleet host, and every other fleet component — the runner, sccache, RustFS, Zot, GARM and the collector — is a digest-pinned binary under systemd. Adding a container runtime to a host purely to run one service would widen its attack surface for no gain.
type Transport ¶
type Transport struct {
Protocol string `json:"protocol" yaml:"protocol"`
SourceAddress string `json:"source_address" yaml:"source_address"`
TargetAddress string `json:"target_address" yaml:"target_address"`
TargetPort int `json:"target_port" yaml:"target_port"`
QueueDirectory string `json:"queue_directory" yaml:"queue_directory"`
Durability Durability `json:"durability" yaml:"durability"`
}
Transport pins how the collector reaches the store. Both hosts sit in one private cloud network, so telemetry never leaves it and the listener is bound to that address rather than to every interface.
type ValidationError ¶
type ValidationError struct{ Issues []Issue }
func (*ValidationError) Error ¶
func (e *ValidationError) Error() string
Source Files