Fix early exit in compatibility check(#41)

.changes/unreleased/Fixed-20241105-192850.yaml

@@ -0,0 +1,3 @@
+kind: Fixed
+body: ydbops now properly continues the restart loop even if listing nodes during maintenance check fails with "retry exceeded" error
+time: 2024-11-05T19:28:50.308019908+01:00

pkg/rolling/rolling.go

@@ -2,6 +2,7 @@ package rolling
 
 import (
 	"bytes"
+	"errors"
 	"fmt"
 	"math"
 	"strings"
@@ -220,6 +221,7 @@ func (r *Rolling) cmsWaitingLoop(task cms.MaintenanceTask, totalNodes int) error
 		task, err = r.cms.RefreshMaintenanceTask(taskID)
 		if err != nil {
 			r.logger.Warnf("Failed to refresh maintenance task: %+v", err)
+			continue
 		}
 
 		// NOTE: compatibility check will not fire if rolling restart just
@@ -235,7 +237,10 @@ func (r *Rolling) cmsWaitingLoop(task cms.MaintenanceTask, totalNodes int) error
 		// issues once more. We better exit quickly.
 		if !r.opts.SuppressCompatibilityCheck {
 			incompatible := r.tryDetectCompatibilityIssues()
-			if incompatible != nil {
+
+			// if error is retryExceeded, just keep trying - maybe you have been asking CMS
+			// from a node that has just been restarted, and it's okay.
+			if incompatible != nil && !errors.Is(incompatible, &utils.RetryExceededError{}) {
 				return incompatible
 			}
 		}

pkg/utils/retries.go

@@ -11,6 +11,24 @@ import (
 	"google.golang.org/grpc/status"
 )
 
+type RetryExceededError struct {
+	msg string
+	err error
+}
+
+func (e *RetryExceededError) Error() string {
+	return e.msg + ". Last error:" + e.err.Error()
+}
+
+func (e *RetryExceededError) Unwrap() error {
+	return e.err
+}
+
+func (e *RetryExceededError) Is(targetErr error) bool {
+	_, ok := targetErr.(*RetryExceededError)
+	return ok
+}
+
 func backoffTimeAfter(attempt int) time.Duration {
 	return time.Second * time.Duration(int(math.Pow(2, float64(attempt))))
 }
@@ -45,5 +63,8 @@ func WrapWithRetries(
 		}
 	}
 
-	return nil, fmt.Errorf("number of retries exceeded: %v. Last error: %w", maxAttempts, lastError)
+	return nil, &RetryExceededError{
+		msg: fmt.Sprintf("number of retries exceeded: %v", maxAttempts),
+		err: lastError,
+	}
 }