Add post_plan_delay option

.github/workflows/goreleaser.yml

@@ -15,12 +15,12 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v4
         with:
-          go-version: "^1.22"
+          go-version: "^1.23"
       - name: Run GoReleaser
         uses: goreleaser/goreleaser-action@v5
         with:
           distribution: goreleaser
           version: latest
           args: release --clean
         env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/test.yml

@@ -10,7 +10,7 @@ jobs:
     strategy:
       matrix:
         go-version:
-          - 1.22.x
+          - 1.23.x
     steps:
       - name: checkout
         uses: actions/checkout@v4
@@ -21,4 +21,4 @@ jobs:
       - name: vet
         run: go vet ./...
       - name: test
-        run: go test -v -race ./...
+        run: go test -v -race ./...

README.md

@@ -44,6 +44,9 @@ groups:
     # How often should we talk to Hetzner to look for changes w.r.t. the infra itself?
     poll_interval: 60s
 
+    # How long to wait after applying a plan before allowing another plan. Default is 0 (disabled).
+    post_plan_delay: 1s
+
     hetzner:
       api_token: "abc123" # Your Hetzner API token.
       project_id: 123456 # Your Hetzner's project ID (you can find it in the URL in the Hetzner dashboard).

config/cfgmodel/model.go

@@ -30,6 +30,11 @@ type GroupConfig struct {
 	// This defaults to 30 seconds.
 	PlanApplyTimeout time.Duration `koanf:"plan_apply_timeout"`
 
+	// PostPlanDelay is the time to wait after a plan is applied before allowing another plan.
+	// This is useful to prevent plans happening too quickly in succession.
+	// This defaults to 0 (disabled).
+	PostPlanDelay time.Duration `koanf:"post_plan_delay"`
+
 	// PlanApplyWithUnknownStatus is a flag that indicates that the group should apply plans
 	// even if the status of one or more servers is unknown.
 	PlanApplyWithUnkownStatus bool `koanf:"plan_apply_with_unknown_status"`
@@ -90,6 +95,11 @@ func (c GroupConfig) PlanApplyTimeoutOrDefault() time.Duration {
 	return c.PlanApplyTimeout
 }
 
+// PostPlanDelayOrDefault returns the post-plan delay (default is 0, disabled).
+func (c GroupConfig) PostPlanDelayOrDefault() time.Duration {
+	return c.PostPlanDelay
+}
+
 // ServiceConfig describes the service. The name is used in metrics and logs.
 type ServiceConfig struct {
 	Name string `koanf:"name"`

go.mod

@@ -1,6 +1,6 @@
 module github.com/gzuidhof/flipper
 
-go 1.22
+go 1.23
 
 require github.com/go-ozzo/ozzo-validation/v4 v4.3.0
 

monitor/group.go

@@ -91,17 +91,20 @@ func (g *Group) executePlan(ctx context.Context, logger *slog.Logger, state plan
 		)
 	}
 
-	if g.provider.Name() == "hetzner" {
-		// Hetzner keeps the floating IPs locked for a short while after assigning them.
-		// So we add a small sleep here to prevent a potential plan that happens right after from failing.
-		// This is a bit of a hack, but it's the simplest solution for now. A potential future solution
-		// could be to retry on lock errors (within the Hetzner provider perhaps).
-		time.Sleep(time.Second)
-	}
-
 	return nil
 }
 
+func (g *Group) applyPostPlanDelay(ctx context.Context) {
+	delay := g.cfg.PostPlanDelayOrDefault()
+	if delay <= 0 {
+		return
+	}
+	select {
+	case <-ctx.Done():
+	case <-time.After(delay):
+	}
+}
+
 // Start watching the resources and performing health checks.
 // This function blocks until the context is cancelled.
 func (g *Group) Start(ctx context.Context) error {
@@ -201,6 +204,8 @@ func (g *Group) Start(ctx context.Context) error {
 				_ = g.notifier.Notify(ctx, msg)
 				logger.InfoContext(ctx, "Plan executed successfully.",
 					slog.Int("num_unhealthy_servers", numUnhealthy))
+
+				g.applyPostPlanDelay(ctx)
 			}
 
 			minSequence = g.watcher.performUpdate(ctx, updateChan, errChan, true)