Fix: Bulk processor retries indefinitely on failure

When all retries are exhausted the worker internal requests buffer needs to be
cleared in failure scenarios. That is required because the commitFunc (and
consequently the underlying BulkService.Do call) doesn't reset it when some error
happens. Without clearing the internal buffer the worker will continue sending
the same requests on the following rounds of execution.

Kudos for this solution goes to @rwynn and @raiRaiyan .

Resolves olivere#1278

bulk_processor.go

@@ -580,6 +580,14 @@ func (w *bulkWorker) commit(ctx context.Context) error {
 	err := RetryNotify(commitFunc, w.p.backoff, notifyFunc)
 	w.updateStats(res)
 	if err != nil {
+		// After all retry attempts clear the requests for the next round. This is
+		// important when backoff is disabled or limited to a number of rounds, and
+		// the aftermath is a failure., because the commitFunc (and consequently the
+		// underlying BulkService.Do call) will not clear the requests from the
+		// worker. Without this the same requests will be used on the next round of
+		// execution of the worker.
+		w.service.Reset()
+
 		w.p.c.errorf("elastic: bulk processor %q failed: %v", w.p.name, err)
 	}
 

bulk_processor_test.go

@@ -8,6 +8,8 @@ import (
 	"context"
 	"fmt"
 	"math/rand"
+	"net/http"
+	"net/http/httptest"
 	"reflect"
 	"sync/atomic"
 	"testing"
@@ -335,6 +337,60 @@ func TestBulkProcessorFlush(t *testing.T) {
 	}
 }
 
+func TestBulkWorker_commit_clearFailedRequests(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
+	defer server.Close()
+
+	client, err := NewClient(
+		SetURL(server.URL),
+		SetSniff(false),
+		SetHealthcheck(false),
+		SetHttpClient(tooManyHTTPClient{}),
+	)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	var calls int
+
+	bulkProcessor, err := NewBulkProcessorService(client).
+		BulkActions(-1).
+		BulkSize(-1).
+		FlushInterval(0).
+		RetryItemStatusCodes().
+		Backoff(StopBackoff{}).
+		After(BulkAfterFunc(func(executionId int64, requests []BulkableRequest, response *BulkResponse, err error) {
+			calls++
+			if calls == 1 {
+				if len(requests) != 10 {
+					t.Errorf("expected 10 requests; got: %d", len(requests))
+				}
+			} else if len(requests) > 0 {
+				t.Errorf("expected 0 requests; got: %d", len(requests))
+			}
+		})).Do(context.Background())
+
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	for i := 0; i < 10; i++ {
+		bulkProcessor.Add(NewBulkIndexRequest())
+	}
+
+	// first flush should process 10 items
+	if err := bulkProcessor.Flush(); err != nil {
+		t.Fatal(err)
+	}
+	// second flush should process none (even if the first flush failed)
+	if err := bulkProcessor.Flush(); err != nil {
+		t.Fatal(err)
+	}
+	if err := bulkProcessor.Close(); err != nil {
+		t.Fatal(err)
+	}
+}
+
 // -- Helper --
 
 func testBulkProcessor(t *testing.T, numDocs int, svc *BulkProcessorService) {
@@ -427,3 +483,12 @@ func testBulkProcessor(t *testing.T, numDocs int, svc *BulkProcessorService) {
 		t.Fatalf("expected %d documents; got: %d", numDocs, count)
 	}
 }
+
+type tooManyHTTPClient struct {
+}
+
+func (t tooManyHTTPClient) Do(r *http.Request) (*http.Response, error) {
+	recorder := httptest.NewRecorder()
+	recorder.WriteHeader(http.StatusTooManyRequests)
+	return recorder.Result(), nil
+}