feat: ability to disable in-sync healthcheck

strangelove-ventures · Oct 8, 2024 · 3c3c9f1 · 3c3c9f1
1 parent f0d8cb8
commit 3c3c9f1
Show file tree

Hide file tree

Showing 3 changed files with 46 additions and 29 deletions.
diff --git a/api/v1/cosmosfullnode_types.go b/api/v1/cosmosfullnode_types.go
@@ -727,6 +727,11 @@ type ServiceSpec struct {
 	// +optional
 	P2PTemplate ServiceOverridesSpec `json:"p2pTemplate"`
 
+	// Disables the health check requiring the instance to be
+	// in-sync with the chain before the pod is marked as ready.
+	// +optional
+	DisableSyncReadiness *bool `json:"disableSyncReadiness"`
+
 	// Overrides for the single RPC service.
 	// +optional
 	RPCTemplate ServiceOverridesSpec `json:"rpcTemplate"`

diff --git a/api/v1/zz_generated.deepcopy.go b/api/v1/zz_generated.deepcopy.go
diff --git a/internal/fullnode/pod_builder.go b/internal/fullnode/pod_builder.go
@@ -90,27 +90,30 @@ func NewPodBuilder(crd *cosmosv1.CosmosFullNode) PodBuilder {
 					ImagePullPolicy: tpl.ImagePullPolicy,
 					WorkingDir:      workDir,
 				},
-				// healthcheck sidecar
-				{
-					Name: "healthcheck",
-					// Available images: https://github.com/orgs/strangelove-ventures/packages?repo_name=cosmos-operator
-					// IMPORTANT: Must use v0.6.2 or later.
-					Image:   "ghcr.io/strangelove-ventures/cosmos-operator:" + version.DockerTag(),
-					Command: []string{"/manager", "healthcheck"},
-					Ports:   []corev1.ContainerPort{{ContainerPort: healthCheckPort, Protocol: corev1.ProtocolTCP}},
-					Resources: corev1.ResourceRequirements{
-						Requests: corev1.ResourceList{
-							corev1.ResourceCPU:    resource.MustParse("5m"),
-							corev1.ResourceMemory: resource.MustParse("16Mi"),
-						},
-					},
-					ReadinessProbe:  probes[1],
-					ImagePullPolicy: tpl.ImagePullPolicy,
-				},
 			},
 		},
 	}
 
+	if crd.Spec.Service.DisableSyncReadiness == nil || !*crd.Spec.Service.DisableSyncReadiness {
+		// healthcheck sidecar
+		pod.Spec.Containers = append(pod.Spec.Containers, corev1.Container{
+			Name: "sync-healthcheck",
+			// Available images: https://github.com/orgs/strangelove-ventures/packages?repo_name=cosmos-operator
+			// IMPORTANT: Must use v0.6.2 or later.
+			Image:   "ghcr.io/strangelove-ventures/cosmos-operator:" + version.DockerTag(),
+			Command: []string{"/manager", "healthcheck"},
+			Ports:   []corev1.ContainerPort{{ContainerPort: healthCheckPort, Protocol: corev1.ProtocolTCP}},
+			Resources: corev1.ResourceRequirements{
+				Requests: corev1.ResourceList{
+					corev1.ResourceCPU:    resource.MustParse("5m"),
+					corev1.ResourceMemory: resource.MustParse("16Mi"),
+				},
+			},
+			ReadinessProbe:  probes[1],
+			ImagePullPolicy: tpl.ImagePullPolicy,
+		})
+	}
+
 	if len(crd.Spec.ChainSpec.Versions) > 0 {
 		// version check sidecar, runs on inverval in case the instance is halting for upgrade.
 		pod.Spec.Containers = append(pod.Spec.Containers, corev1.Container{
@@ -159,19 +162,23 @@ func podReadinessProbes(crd *cosmosv1.CosmosFullNode) []*corev1.Probe {
 		FailureThreshold:    5,
 	}
 
-	sidecarProbe := &corev1.Probe{
-		ProbeHandler: corev1.ProbeHandler{
-			HTTPGet: &corev1.HTTPGetAction{
-				Path:   "/",
-				Port:   intstr.FromInt(healthCheckPort),
-				Scheme: corev1.URISchemeHTTP,
+	var sidecarProbe *corev1.Probe
+
+	if crd.Spec.Service.DisableSyncReadiness == nil || !*crd.Spec.Service.DisableSyncReadiness {
+		sidecarProbe = &corev1.Probe{
+			ProbeHandler: corev1.ProbeHandler{
+				HTTPGet: &corev1.HTTPGetAction{
+					Path:   "/",
+					Port:   intstr.FromInt(healthCheckPort),
+					Scheme: corev1.URISchemeHTTP,
+				},
 			},
-		},
-		InitialDelaySeconds: 1,
-		TimeoutSeconds:      10,
-		PeriodSeconds:       10,
-		SuccessThreshold:    1,
-		FailureThreshold:    3,
+			InitialDelaySeconds: 1,
+			TimeoutSeconds:      10,
+			PeriodSeconds:       10,
+			SuccessThreshold:    1,
+			FailureThreshold:    3,
+		}
 	}
 
 	return []*corev1.Probe{mainProbe, sidecarProbe}