Less restrictive approach to the replay, booting Oracle even if the r…

…eplay fails (#747)

core/services/ocr2/plugins/ccip/internal/oraclelib/backfilled_oracle.go

@@ -74,13 +74,12 @@ func (r *BackfilledOracle) Run() {
 			errMu.Lock()
 			err = multierr.Combine(err, dstReplayErr)
 			errMu.Unlock()
-			r.lggr.Infow("finished replaying  dst chain", "time", time.Since(s))
+			r.lggr.Infow("finished replaying dst chain", "time", time.Since(s))
 		}()
 	}
 	wg.Wait()
 	if err != nil {
-		r.lggr.Errorw("unexpected error replaying", "err", err)
-		return
+		r.lggr.Criticalw("unexpected error replaying, continuing plugin boot without all the logs backfilled", "err", err)
 	}
 	if err := ctx.Err(); err != nil {
 		r.lggr.Errorw("context already cancelled", "err", err)

core/services/ocr2/plugins/ccip/internal/oraclelib/backfilled_oracle_test.go

@@ -42,14 +42,15 @@ func TestBackfilledOracle(t *testing.T) {
 	job2.Run()
 	assert.True(t, job2.IsRunning())
 
-	/// Replay fails, so no Start()
+	/// Replay fails, but it starts anyway
 	lp11 := lpmocks.NewLogPoller(t)
 	lp12 := lpmocks.NewLogPoller(t)
 	lp11.On("Replay", mock.Anything, int64(1)).Return(errors.New("Replay failed")).Once()
 	lp12.On("Replay", mock.Anything, int64(2)).Return(errors.New("Replay failed")).Once()
 
 	oracle := jobmocks.NewServiceCtx(t)
+	oracle.On("Start", mock.Anything).Return(nil).Once()
 	job3 := NewBackfilledOracle(logger.NullLogger, lp11, lp12, 1, 2, oracle)
 	job3.Run()
-	assert.False(t, job3.IsRunning())
+	assert.True(t, job3.IsRunning())
 }