Merge pull request #201 from HarrisonWAffel/retries-after-restart

Add ResetFailureCountOnServiceRestart

pkg/applyinator/applyinator.go

@@ -45,6 +45,9 @@ type Plan struct {
 	OneTimeInstructions  []OneTimeInstruction    `json:"instructions,omitempty"`
 	Probes               map[string]prober.Probe `json:"probes,omitempty"`
 	PeriodicInstructions []PeriodicInstruction   `json:"periodicInstructions,omitempty"`
+	// ResetFailureCountOnStartup denotes whether the system-agent should reset the failure count
+	// and applied-checksum for plans that are force applied each time the system-agent starts.
+	ResetFailureCountOnStartup bool `json:"resetFailureCountOnStartup,omitempty"`
 }
 
 type CommonInstruction struct {

pkg/k8splan/watcher.go

@@ -215,6 +215,12 @@ func (w *watcher) start(ctx context.Context, strictVerify bool) {
 				logrus.Infof("Detected first start, force-applying one-time instruction set")
 				needsApplied = true
 				hasRunOnce = true
+				// Plans which have previously succeeded but need to be force applied
+				// should continue to respect the specified failure count.
+				if cp.Plan.ResetFailureCountOnStartup {
+					secret.Data[appliedChecksumKey] = []byte("")
+					secret.Data[failureCountKey] = []byte("0")
+				}
 			}
 
 			// Check to see if we've exceeded our failure count threshold