Expose a single timeout setting in CRDs (#1045)

Signed-off-by: Pavol Loffay <[email protected]>
grafana · Oct 7, 2024 · c7eb01b · c7eb01b
1 parent e2ae4c0
commit c7eb01b
Show file tree

Hide file tree

Showing 51 changed files with 417 additions and 78 deletions.
diff --git a/.chloggen/timeout.yaml b/.chloggen/timeout.yaml
@@ -0,0 +1,18 @@
+# One of 'breaking', 'deprecation', 'new_component', 'enhancement', 'bug_fix'
+change_type: breaking
+
+# The name of the component, or a single word describing the area of concern, (e.g. tempostack, tempomonolithic, github action)
+component: tempostack, tempomonolithic
+
+# A brief description of the change. Surround your text with quotes ("") if it needs to start with a backtick (`).
+note: Add unified timeout configuration. It changes the default to 30s.
+
+# One or more tracking issues related to the change
+issues: [1045]
+
+# (Optional) One or more lines of additional information to render under the primary note.
+# These lines will be padded with 2 spaces and then inserted directly into the document.
+# Use pipe (|) for multiline entries.
+subtext: |
+  Adding `spec.timeout` CRD option to configure timeout on all components and default it to 30s.
+  Before Tempo server was defaulting to 3m, gateway to 2m, OpenShift route to 30s (for query), oauth-proxy to 30s (for query).
diff --git a/Makefile b/Makefile
@@ -6,7 +6,7 @@ TEMPO_QUERY_VERSION ?= main-2999520
 JAEGER_QUERY_VERSION ?= d6631f5f2370cfc3a49efce312491031fb387600
 TEMPO_GATEWAY_VERSION ?= main-2024-08-05-11d0d94
 TEMPO_GATEWAY_OPA_VERSION ?= main-2024-04-29-914c13f
-OAUTH_PROXY_VERSION=4.12
+OAUTH_PROXY_VERSION=4.14
 
 MIN_KUBERNETES_VERSION ?= 1.25.0
 MIN_OPENSHIFT_VERSION ?= 4.12

diff --git a/apis/tempo/v1alpha1/tempomonolithic_defaults.go b/apis/tempo/v1alpha1/tempomonolithic_defaults.go
@@ -16,6 +16,7 @@ var (
 	twoGBQuantity           = resource.MustParse("2Gi")
 	tenGBQuantity           = resource.MustParse("10Gi")
 	defaultServicesDuration = metav1.Duration{Duration: time.Hour * 24 * 3}
+	defaultTimeout          = metav1.Duration{Duration: time.Second * 30}
 )
 
 // Default sets all default values in a central place, instead of setting it at every place where the value is accessed.
@@ -88,4 +89,8 @@ func (r *TempoMonolithic) Default(ctrlConfig configv1alpha1.ProjectConfig) {
 			r.Spec.JaegerUI.ServicesQueryDuration = &defaultServicesDuration
 		}
 	}
+
+	if r.Spec.Timeout.Duration == 0 {
+		r.Spec.Timeout = defaultTimeout
+	}
 }
diff --git a/apis/tempo/v1alpha1/tempomonolithic_defaults_test.go b/apis/tempo/v1alpha1/tempomonolithic_defaults_test.go
@@ -6,6 +6,7 @@ import (
 
 	"github.com/stretchr/testify/assert"
 	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	configv1alpha1 "github.com/grafana/tempo-operator/apis/config/v1alpha1"
@@ -46,6 +47,7 @@ func TestMonolithicDefault(t *testing.T) {
 						},
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -79,6 +81,7 @@ func TestMonolithicDefault(t *testing.T) {
 						},
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -104,6 +107,7 @@ func TestMonolithicDefault(t *testing.T) {
 						},
 					},
 					Management: "Unmanaged",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 			expected: &TempoMonolithic{
@@ -125,6 +129,7 @@ func TestMonolithicDefault(t *testing.T) {
 						},
 					},
 					Management: "Unmanaged",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -194,6 +199,7 @@ func TestMonolithicDefault(t *testing.T) {
 						ServicesQueryDuration: &defaultServicesDuration,
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -266,6 +272,7 @@ func TestMonolithicDefault(t *testing.T) {
 						ServicesQueryDuration: &defaultServicesDuration,
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -330,6 +337,7 @@ func TestMonolithicDefault(t *testing.T) {
 						ServicesQueryDuration: &defaultServicesDuration,
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
@@ -393,11 +401,12 @@ func TestMonolithicDefault(t *testing.T) {
 						ServicesQueryDuration: &defaultServicesDuration,
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Second * 30},
 				},
 			},
 		},
 		{
-			name: "define custom duration for services list",
+			name: "define custom duration for services list and timeout",
 			input: &TempoMonolithic{
 				ObjectMeta: v1.ObjectMeta{
 					Name:      "test",
@@ -417,6 +426,7 @@ func TestMonolithicDefault(t *testing.T) {
 						},
 						ServicesQueryDuration: &v1.Duration{Duration: time.Duration(100 * 100)},
 					},
+					Timeout: metav1.Duration{Duration: time.Hour},
 				},
 			},
 			expected: &TempoMonolithic{
@@ -454,6 +464,7 @@ func TestMonolithicDefault(t *testing.T) {
 						ServicesQueryDuration: &v1.Duration{Duration: time.Duration(100 * 100)},
 					},
 					Management: "Managed",
+					Timeout:    metav1.Duration{Duration: time.Hour},
 				},
 			},
 		},

diff --git a/apis/tempo/v1alpha1/tempomonolithic_types.go b/apis/tempo/v1alpha1/tempomonolithic_types.go
@@ -44,6 +44,11 @@ type TempoMonolithicSpec struct {
 	// +operator-sdk:csv:customresourcedefinitions:type=spec,displayName="Resources",order=5,xDescriptors="urn:alm:descriptor:com.tectonic.ui:resourceRequirements"
 	Resources *corev1.ResourceRequirements `json:"resources,omitempty"`
 
+	// Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+	// Timeout configuration on a specific component has a higher precedence.
+	// Default is 30 seconds.
+	Timeout metav1.Duration `json:"timeout,omitempty"`
+
 	// ServiceAccount defines the Service Account to use for all Tempo components.
 	//
 	// +kubebuilder:validation:Optional

diff --git a/apis/tempo/v1alpha1/tempostack_types.go b/apis/tempo/v1alpha1/tempostack_types.go
@@ -40,6 +40,11 @@ type TempoStackSpec struct {
 	// +operator-sdk:csv:customresourcedefinitions:type=spec,displayName="Ingestion and Querying Ratelimiting"
 	LimitSpec LimitSpec `json:"limits,omitempty"`
 
+	// Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+	// Timeout configuration on a specific component has a higher precedence.
+	// Defaults to 30 seconds.
+	Timeout metav1.Duration `json:"timeout,omitempty"`
+
 	// StorageClassName for PVCs used by ingester. Defaults to nil (default storage class in the cluster).
 	//
 	// +optional

diff --git a/apis/tempo/v1alpha1/zz_generated.deepcopy.go b/apis/tempo/v1alpha1/zz_generated.deepcopy.go
diff --git a/bundle/community/manifests/tempo-operator.clusterserviceversion.yaml b/bundle/community/manifests/tempo-operator.clusterserviceversion.yaml
@@ -74,7 +74,7 @@ metadata:
     capabilities: Deep Insights
     categories: Logging & Tracing,Monitoring
     containerImage: ghcr.io/grafana/tempo-operator/tempo-operator:v0.13.0
-    createdAt: "2024-10-04T12:51:28Z"
+    createdAt: "2024-10-07T07:11:28Z"
     description: Create and manage deployments of Tempo, a high-scale distributed
       tracing backend.
     operatorframework.io/cluster-monitoring: "true"
@@ -1433,7 +1433,7 @@ spec:
                 - name: RELATED_IMAGE_TEMPO_GATEWAY_OPA
                   value: quay.io/observatorium/opa-openshift:main-2024-04-29-914c13f
                 - name: RELATED_IMAGE_OAUTH_PROXY
-                  value: quay.io/openshift/origin-oauth-proxy:4.12
+                  value: quay.io/openshift/origin-oauth-proxy:4.14
                 image: ghcr.io/grafana/tempo-operator/tempo-operator:v0.13.0
                 livenessProbe:
                   httpGet:
@@ -1582,7 +1582,7 @@ spec:
     name: tempo-gateway
   - image: quay.io/observatorium/opa-openshift:main-2024-04-29-914c13f
     name: tempo-gateway-opa
-  - image: quay.io/openshift/origin-oauth-proxy:4.12
+  - image: quay.io/openshift/origin-oauth-proxy:4.14
     name: oauth-proxy
   version: 0.13.0
   webhookdefinitions:

diff --git a/bundle/community/manifests/tempo.grafana.com_tempomonolithics.yaml b/bundle/community/manifests/tempo.grafana.com_tempomonolithics.yaml
@@ -1683,6 +1683,12 @@ spec:
                 required:
                 - traces
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Default is 30 seconds.
+                type: string
               tolerations:
                 description: Tolerations defines the tolerations of a node to schedule
                   the pod onto it.

diff --git a/bundle/community/manifests/tempo.grafana.com_tempostacks.yaml b/bundle/community/manifests/tempo.grafana.com_tempostacks.yaml
@@ -2715,6 +2715,12 @@ spec:
                 required:
                 - mode
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Defaults to 30 seconds.
+                type: string
             required:
             - storage
             type: object

diff --git a/bundle/openshift/manifests/tempo-operator.clusterserviceversion.yaml b/bundle/openshift/manifests/tempo-operator.clusterserviceversion.yaml
@@ -74,7 +74,7 @@ metadata:
     capabilities: Deep Insights
     categories: Logging & Tracing,Monitoring
     containerImage: ghcr.io/grafana/tempo-operator/tempo-operator:v0.13.0
-    createdAt: "2024-10-04T12:51:27Z"
+    createdAt: "2024-10-07T07:11:27Z"
     description: Create and manage deployments of Tempo, a high-scale distributed
       tracing backend.
     operatorframework.io/cluster-monitoring: "true"
@@ -1443,7 +1443,7 @@ spec:
                 - name: RELATED_IMAGE_TEMPO_GATEWAY_OPA
                   value: quay.io/observatorium/opa-openshift:main-2024-04-29-914c13f
                 - name: RELATED_IMAGE_OAUTH_PROXY
-                  value: quay.io/openshift/origin-oauth-proxy:4.12
+                  value: quay.io/openshift/origin-oauth-proxy:4.14
                 image: ghcr.io/grafana/tempo-operator/tempo-operator:v0.13.0
                 livenessProbe:
                   httpGet:
@@ -1603,7 +1603,7 @@ spec:
     name: tempo-gateway
   - image: quay.io/observatorium/opa-openshift:main-2024-04-29-914c13f
     name: tempo-gateway-opa
-  - image: quay.io/openshift/origin-oauth-proxy:4.12
+  - image: quay.io/openshift/origin-oauth-proxy:4.14
     name: oauth-proxy
   version: 0.13.0
   webhookdefinitions:

diff --git a/bundle/openshift/manifests/tempo.grafana.com_tempomonolithics.yaml b/bundle/openshift/manifests/tempo.grafana.com_tempomonolithics.yaml
@@ -1683,6 +1683,12 @@ spec:
                 required:
                 - traces
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Default is 30 seconds.
+                type: string
               tolerations:
                 description: Tolerations defines the tolerations of a node to schedule
                   the pod onto it.

diff --git a/bundle/openshift/manifests/tempo.grafana.com_tempostacks.yaml b/bundle/openshift/manifests/tempo.grafana.com_tempostacks.yaml
@@ -2715,6 +2715,12 @@ spec:
                 required:
                 - mode
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Defaults to 30 seconds.
+                type: string
             required:
             - storage
             type: object

diff --git a/config/crd/bases/tempo.grafana.com_tempomonolithics.yaml b/config/crd/bases/tempo.grafana.com_tempomonolithics.yaml
@@ -1679,6 +1679,12 @@ spec:
                 required:
                 - traces
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Default is 30 seconds.
+                type: string
               tolerations:
                 description: Tolerations defines the tolerations of a node to schedule
                   the pod onto it.

diff --git a/config/crd/bases/tempo.grafana.com_tempostacks.yaml b/config/crd/bases/tempo.grafana.com_tempostacks.yaml
@@ -2711,6 +2711,12 @@ spec:
                 required:
                 - mode
                 type: object
+              timeout:
+                description: |-
+                  Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier.
+                  Timeout configuration on a specific component has a higher precedence.
+                  Defaults to 30 seconds.
+                type: string
             required:
             - storage
             type: object

diff --git a/config/manager/manager.yaml b/config/manager/manager.yaml
@@ -50,7 +50,7 @@ spec:
         - name: RELATED_IMAGE_TEMPO_GATEWAY_OPA
           value: quay.io/observatorium/opa-openshift:main-2024-04-29-914c13f
         - name: RELATED_IMAGE_OAUTH_PROXY
-          value: quay.io/openshift/origin-oauth-proxy:4.12
+          value: quay.io/openshift/origin-oauth-proxy:4.14
         securityContext:
           allowPrivilegeEscalation: false
           capabilities:

diff --git a/docs/spec/tempo.grafana.com_tempomonolithics.yaml b/docs/spec/tempo.grafana.com_tempomonolithics.yaml
@@ -126,6 +126,7 @@ spec:                                    # TempoMonolithicSpec defines the desir
           certName: ""                   # Cert is the name of a Secret containing a certificate (tls.crt) and private key (tls.key). It needs to be in the same namespace as the Tempo custom resource.
           minVersion: ""                 # MinVersion defines the minimum acceptable TLS version.
       size: 0Gi                          # Size defines the size of the volume where traces are stored. For in-memory storage, this defines the size of the tmpfs volume. For persistent volume storage, this defines the size of the persistent volume. For object storage, this defines the size of the persistent volume containing the Write-Ahead Log (WAL) of Tempo. Default: 2Gi for memory, 10Gi for all other backends.
+  timeout: ""                            # Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier. Timeout configuration on a specific component has a higher precedence. Default is 30 seconds.
   affinity:                              # Affinity defines the Affinity rules for scheduling pods.
     nodeAffinity: {}                     # Describes node affinity scheduling rules for the pod.
     podAffinity: {}                      # Describes pod affinity scheduling rules (e.g. co-locate this pod in the same node, zone, etc. as some other pod(s)).

diff --git a/docs/spec/tempo.grafana.com_tempostacks.yaml b/docs/spec/tempo.grafana.com_tempostacks.yaml
@@ -388,6 +388,7 @@ spec:                                    # TempoStackSpec defines the desired st
         resources:
         - ""
     mode: "static"                       # Mode defines the multitenancy mode.
+  timeout: ""                            # Timeout configures the same timeout on all components starting at ingress down to the ingestor/querier. Timeout configuration on a specific component has a higher precedence. Defaults to 30 seconds.
   resources:                             # Resources defines resources configuration.
     total:                               # The total amount of resources for Tempo instance. The operator autonomously splits resources between deployed Tempo components. Only limits are supported, the operator calculates requests automatically. See http://github.com/grafana/tempo/issues/1540.
       claims:                            # Claims lists the names of resources, defined in spec.resourceClaims, that are used by this container.   This is an alpha field and requires enabling the DynamicResourceAllocation feature gate.   This field is immutable. It can only be set for containers.

diff --git a/internal/manifests/config/build.go b/internal/manifests/config/build.go
@@ -89,6 +89,7 @@ func buildConfiguration(params manifestutils.Params) ([]byte, error) {
 		TLS:          tlsopts,
 		ReceiverTLS:  buildReceiverTLSConfig(tempo),
 		S3StorageTLS: buildS3StorageTLSConfig(params),
+		Timeout:      params.Tempo.Spec.Timeout.Duration,
 	}
 
 	if isTenantOverridesConfigRequired(tempo.Spec.LimitSpec) {