Add the health engine: through-proxy probes, thresholds, channel-fed transitions

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
2026-08-09 15:01:10 +02:00
parent 71c00c40d1
commit 801a9fbe5f
10 changed files with 1350 additions and 34 deletions

View File

@@ -2,6 +2,7 @@ package controller
import (
"context"
"fmt"
"k8s.io/apimachinery/pkg/api/equality"
apimeta "k8s.io/apimachinery/pkg/api/meta"
@@ -11,8 +12,9 @@ import (
crawlv1alpha1 "gitea.home.hrajfrisbee.cz/kacerr/egress-proxies-operator/api/v1alpha1"
)
// Reasons used on the Provisioned condition. The Healthy condition is owned
// by the health engine (internal/health) and only represented here.
// Reasons used on the Provisioned condition, plus the two the reconciler
// writes on the Healthy condition when representing the health engine's
// verdict (the engine owns the state; only the reconciler writes status).
const (
ReasonProvisioning = "Provisioning"
ReasonCreated = "Created"
@@ -23,6 +25,9 @@ const (
ReasonCloudInitError = "CloudInitError"
ReasonExternalEndpoint = "ExternalEndpoint"
ReasonDeleting = "Deleting"
ReasonProbeSucceeded = "ProbeSucceeded"
ReasonProbeFailed = "ProbeFailed"
)
// setProvisioned stages the Provisioned condition on p. Nothing is written
@@ -39,6 +44,39 @@ func setProvisioned(p *crawlv1alpha1.Proxy, status metav1.ConditionStatus, reaso
})
}
// applyHealth stages the Healthy condition and the latency fields from the
// health engine's current snapshot. Called only from states where the proxy
// is reachable (Running, External); everywhere else the condition is either
// left as-is or removed by the create branch.
func (r *ProxyReconciler) applyHealth(p *crawlv1alpha1.Proxy) {
if r.Health == nil {
return
}
snap, ok := r.Health.Snapshot(client.ObjectKeyFromObject(p))
if !ok {
return
}
cond := metav1.Condition{
Type: crawlv1alpha1.ConditionHealthy,
ObservedGeneration: p.Generation,
}
if snap.Healthy {
cond.Status = metav1.ConditionTrue
cond.Reason = ReasonProbeSucceeded
cond.Message = "probe succeeded through the proxy"
} else {
cond.Status = metav1.ConditionFalse
cond.Reason = ReasonProbeFailed
cond.Message = fmt.Sprintf("%d consecutive probe failures; last: %s",
snap.ConsecutiveFailures, snap.LastError)
}
apimeta.SetStatusCondition(&p.Status.Conditions, cond)
p.Status.LatencyMillis = snap.Latency.Milliseconds()
if !snap.LastProbe.IsZero() {
p.Status.LastHealthCheckTime = &metav1.Time{Time: snap.LastProbe}
}
}
// computePhase derives status.phase from deletionTimestamp and the
// Provisioned/Healthy conditions. Pure, so the truth table is unit-testable.
func computePhase(p *crawlv1alpha1.Proxy) crawlv1alpha1.ProxyPhase {