Retry DonsForCapability during cap init

chain_capabilities/common/transmission_schedule/transmission_schedule.go

@@ -15,6 +15,8 @@ import (
 	"github.com/smartcontractkit/chainlink-common/pkg/capabilities"
 	"github.com/smartcontractkit/chainlink-common/pkg/logger"
 	"github.com/smartcontractkit/chainlink-common/pkg/types/core"
+
+	capcommon "github.com/smartcontractkit/capabilities/chain_capabilities/common"
 )
 
 // TransmissionScheduler handles capability-layer transmission scheduling by waiting for the appropriate delay
@@ -118,7 +120,15 @@ func InitMyDON(ctx context.Context, registry core.CapabilitiesRegistry, capabili
 		return capabilities.DON{}, fmt.Errorf("capabilities registry is nil")
 	}
 
-	donsWithNodes, err := registry.DONsForCapability(ctx, capabilityID)
+	// The host node populates its capabilities metadata registry asynchronously
+	// (CapabilitiesLauncher.OnNewRegistry after the first on-chain registry sync).
+	// A plugin's Initialise can win the race and call in before that happens, in
+	// which case the registry returns "metadataRegistry information not available".
+	// Retry with backoff until the registry is ready or the host-provided ctx
+	// (bounded by the standard-capabilities startTimeout) expires.
+	donsWithNodes, err := capcommon.WithPollingRetry(ctx, lggr, func(ctx context.Context) ([]capabilities.DONWithNodes, error) {
+		return registry.DONsForCapability(ctx, capabilityID)
+	})
 	if err != nil {
 		lggr.Errorw("failed getting DONs for capability", "capabilityID", capabilityID, "error", err)
 		return capabilities.DON{}, fmt.Errorf("failed getting dons for capability: %w", err)
@@ -139,6 +149,9 @@ func InitMyDON(ctx context.Context, registry core.CapabilitiesRegistry, capabili
 	}
 
 	// Legacy path: filter by local PeerID to find which DON(s) this node belongs to.
+	// No retry needed here: DONsForCapability above already gates on the metadata
+	// registry being populated, so by this point LocalNode will not hit the
+	// "metadataRegistry information not available" race.
 	localNode, err := registry.LocalNode(ctx)
 	if err != nil {
 		lggr.Errorw("failed to get local node", "error", err)

chain_capabilities/common/transmission_schedule/transmission_schedule_test.go

@@ -2,6 +2,7 @@ package transmissionschedule_test
 
 import (
 	"context"
+	"errors"
 	"testing"
 	"time"
 
@@ -183,6 +184,19 @@ func TestInitMyDON(t *testing.T) {
 		require.EqualValues(t, 20, got.ID)
 	})
 
+	t.Run("authoritative: returns correct DON by ID after failing once", func(t *testing.T) {
+		reg := mocks.NewCapabilitiesRegistry(t)
+		reg.EXPECT().DONsForCapability(mock.Anything, capID).Return(nil, errors.New("metadataRegistry information not available")).Once()
+		reg.EXPECT().DONsForCapability(mock.Anything, capID).Return(buildDONs(map[uint32][]p2ptypes.PeerID{
+			10: {myPeerID, otherPeerID},
+			20: {myPeerID, otherPeerID},
+		}), nil)
+
+		got, err := ts.InitMyDON(context.Background(), reg, capID, 20, logger.Test(t), false)
+		require.NoError(t, err)
+		require.EqualValues(t, 20, got.ID)
+	})
+
 	t.Run("authoritative: selects correct DON even when local node is not listed as a member of all DONs", func(t *testing.T) {
 		// Plex scenario: node serves DON 20 but the registry lists it only in DON 10's
 		// membership. With the authoritative ID we bypass peer-membership filtering
@@ -234,6 +248,19 @@ func TestInitMyDON(t *testing.T) {
 		require.EqualValues(t, 10, got.ID)
 	})
 
+	t.Run("legacy: single matched DON returns that DON after failing once", func(t *testing.T) {
+		reg := mocks.NewCapabilitiesRegistry(t)
+		withLocalNode(reg)
+		reg.EXPECT().DONsForCapability(mock.Anything, capID).Return(nil, errors.New("metadataRegistry information not available")).Once()
+		reg.EXPECT().DONsForCapability(mock.Anything, capID).Return(buildDONs(map[uint32][]p2ptypes.PeerID{
+			10: {myPeerID, otherPeerID},
+		}), nil)
+
+		got, err := ts.InitMyDON(context.Background(), reg, capID, 0, logger.Test(t), false)
+		require.NoError(t, err)
+		require.EqualValues(t, 10, got.ID)
+	})
+
 	t.Run("legacy: multiple matched DONs returns first and warns", func(t *testing.T) {
 		reg := mocks.NewCapabilitiesRegistry(t)
 		withLocalNode(reg)