chore: remove max retries

flanksource · Apr 18, 2024 · 25c4266 · 25c4266
1 parent 77c8f2a
commit 25c4266
Show file tree

Hide file tree

Showing 10 changed files with 14 additions and 49 deletions.
diff --git a/api/v1/checks.go b/api/v1/checks.go
@@ -790,10 +790,9 @@ type KubernetesResourceChecks struct {
 
 type KubernetesResourceCheckRetries struct {
 	// Delay is the initial delay
-	Delay      string `json:"delay,omitempty"`
-	Timeout    string `json:"timeout,omitempty"`
-	Interval   string `json:"interval,omitempty"`
-	MaxRetries int    `json:"maxRetries,omitempty"`
+	Delay    string `json:"delay,omitempty"`
+	Timeout  string `json:"timeout,omitempty"`
+	Interval string `json:"interval,omitempty"`
 
 	parsedDelay    *time.Duration `json:"-"`
 	parsedTimeout  *time.Duration `json:"-"`
@@ -871,11 +870,9 @@ type KubernetesResourceCheckWaitFor struct {
 	Timeout string `json:"timeout,omitempty"`
 
 	// Interval to check if all static & non-static resources are ready.
-	// 	Default: 30s
+	// 	Default: 5s
 	Interval string `json:"interval,omitempty"`
 
-	MaxRetries int `json:"maxRetries,omitempty"`
-
 	parsedTimeout  *time.Duration `json:"-"`
 	parsedInterval *time.Duration `json:"-"`
 }

diff --git a/checks/kubernetes_resource.go b/checks/kubernetes_resource.go
@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"strconv"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/flanksource/gomplate/v3"
@@ -15,7 +16,6 @@ import (
 	apiErrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
-	"k8s.io/apimachinery/pkg/runtime/schema"
 	cliresource "k8s.io/cli-runtime/pkg/resource"
 
 	"github.com/flanksource/canary-checker/api/context"
@@ -145,10 +145,6 @@ func (c *KubernetesResourceChecker) Check(ctx *context.Context, check v1.Kuberne
 			backoff = retry.NewConstant(retryInterval)
 		}
 
-		if check.CheckRetries.MaxRetries > 0 {
-			backoff = retry.WithMaxRetries(uint64(check.CheckRetries.MaxRetries), backoff)
-		}
-
 		if maxRetryTimeout, _ := check.CheckRetries.GetTimeout(); maxRetryTimeout > 0 {
 			backoff = retry.WithMaxDuration(maxRetryTimeout, backoff)
 		}
@@ -198,9 +194,6 @@ func (c *KubernetesResourceChecker) evalWaitFor(ctx *context.Context, check v1.K
 
 	var attempts int
 	backoff := retry.WithMaxDuration(waitTimeout, retry.NewConstant(waitInterval))
-	if check.WaitFor.MaxRetries > 0 {
-		backoff = retry.WithMaxRetries(uint64(check.WaitFor.MaxRetries), backoff)
-	}
 	retryErr := retry.Do(ctx, backoff, func(_ctx gocontext.Context) error {
 		ctx = _ctx.(*context.Context)
 		attempts++
@@ -299,7 +292,7 @@ func deleteResources(ctx *context.Context, waitForDelete bool, resources ...unst
 	ctx.Logger.V(4).Infof("deleting %d resources", len(resources))
 
 	// cache dynamic clients
-	clients := map[schema.GroupVersionKind]*cliresource.Helper{}
+	clients := sync.Map{}
 
 	eg, _ := errgroup.WithContext(ctx)
 	for i := range resources {
@@ -311,7 +304,7 @@ func deleteResources(ctx *context.Context, waitForDelete bool, resources ...unst
 				return fmt.Errorf("failed to get rest client for (%s/%s/%s): %w", resource.GetKind(), resource.GetNamespace(), resource.GetName(), err)
 			}
 			gvk := resource.GetObjectKind().GroupVersionKind()
-			clients[gvk] = rc
+			clients.Store(gvk, rc)
 
 			namespace := utils.Coalesce(resource.GetNamespace(), ctx.Namespace)
 			deleteOpt := &metav1.DeleteOptions{
@@ -354,7 +347,9 @@ func deleteResources(ctx *context.Context, waitForDelete bool, resources ...unst
 
 			deleted := make(map[string]struct{})
 			for _, resource := range resources {
-				rc := clients[resource.GetObjectKind().GroupVersionKind()]
+				cachedClient, _ := clients.Load(resource.GetObjectKind().GroupVersionKind())
+				rc := cachedClient.(*cliresource.Helper)
+
 				if _, err := rc.Get(resource.GetNamespace(), resource.GetName()); err != nil {
 					if !apiErrors.IsNotFound(err) {
 						return fmt.Errorf("error getting resource (%s/%s/%s) while polling: %w", resource.GetKind(), resource.GetNamespace(), resource.GetName(), err)

diff --git a/config/deploy/crd.yaml b/config/deploy/crd.yaml
@@ -4997,8 +4997,6 @@ spec:
                             type: string
                           interval:
                             type: string
-                          maxRetries:
-                            type: integer
                           timeout:
                             type: string
                         type: object
@@ -5163,10 +5161,8 @@ spec:
                           interval:
                             description: |-
                               Interval to check if all static & non-static resources are ready.
-                              	Default: 30s
+                              	Default: 5s
                             type: string
-                          maxRetries:
-                            type: integer
                           timeout:
                             description: |-
                               Timeout to wait for all static & non-static resources to be ready.

diff --git a/config/deploy/manifests.yaml b/config/deploy/manifests.yaml
@@ -4996,8 +4996,6 @@ spec:
                             type: string
                           interval:
                             type: string
-                          maxRetries:
-                            type: integer
                           timeout:
                             type: string
                         type: object
@@ -5162,10 +5160,8 @@ spec:
                           interval:
                             description: |-
                               Interval to check if all static & non-static resources are ready.
-                              	Default: 30s
+                              	Default: 5s
                             type: string
-                          maxRetries:
-                            type: integer
                           timeout:
                             description: |-
                               Timeout to wait for all static & non-static resources to be ready.

diff --git a/config/schemas/canary.schema.json b/config/schemas/canary.schema.json
@@ -2353,9 +2353,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,
@@ -2377,9 +2374,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,

diff --git a/config/schemas/component.schema.json b/config/schemas/component.schema.json
@@ -2610,9 +2610,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,
@@ -2634,9 +2631,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,

diff --git a/config/schemas/topology.schema.json b/config/schemas/topology.schema.json
@@ -2580,9 +2580,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,
@@ -2604,9 +2601,6 @@
         },
         "interval": {
           "type": "string"
-        },
-        "maxRetries": {
-          "type": "integer"
         }
       },
       "additionalProperties": false,

diff --git a/fixtures/k8s/kubernetes_resource_ingress_pass.yaml b/fixtures/k8s/kubernetes_resource_ingress_pass.yaml
@@ -72,4 +72,4 @@ spec:
       checkRetries:
         delay: 3s
         interval: 2s
-        maxRetries: 5
+        timeout: 5m
diff --git a/fixtures/k8s/kubernetes_resource_pod_exit_code_pass.yaml b/fixtures/k8s/kubernetes_resource_pod_exit_code_pass.yaml
@@ -29,7 +29,7 @@ spec:
         timeout: "20s"
       checkRetries:
         delay: 2s
-        maxRetries: 3
+        timeout: 5m
       checks:
         - kubernetes:
             - name: exit-code-check

diff --git a/fixtures/k8s/kubernetes_resource_service_pass.yaml b/fixtures/k8s/kubernetes_resource_service_pass.yaml
@@ -43,5 +43,4 @@ spec:
               url: "http://httpbin-svc.default.svc"
       checkRetries:
         delay: 2s
-        maxRetries: 5
         interval: 3s