roachtest/cdc/mixed-versions: allow for changefeed failures

This test may flake due to the upgrade from 22.1->22.2. The
test asserts a changefeed remains running by checking for
resolved timestamps being emitted on a regular basis. The
problem with this is that, during the rolling upgrade,
the changefeed may fail with a "draining" error.
This issue is fixed in 22.2 onwards by treating all errors
as retryable.

Rather than skipping this test because 22.1 is EOLed, it is
preferable to still run this test regularly because it tests
22.2 functionality. This change adds a fix where the test
will poll the changefeed every 1s and recreate it if it fails.

Closes: #106878
Release note: None
Epic: None

pkg/cmd/roachtest/tests/mixed_version_cdc.go

@@ -25,11 +25,13 @@ import (
 	"github.com/cockroachdb/cockroach/pkg/cmd/roachtest/registry"
 	"github.com/cockroachdb/cockroach/pkg/cmd/roachtest/spec"
 	"github.com/cockroachdb/cockroach/pkg/cmd/roachtest/test"
+	"github.com/cockroachdb/cockroach/pkg/jobs"
 	"github.com/cockroachdb/cockroach/pkg/roachprod/vm"
 	"github.com/cockroachdb/cockroach/pkg/util/randutil"
 	"github.com/cockroachdb/cockroach/pkg/util/syncutil"
 	"github.com/cockroachdb/cockroach/pkg/util/timeutil"
 	"github.com/cockroachdb/cockroach/pkg/util/version"
+	"github.com/cockroachdb/errors"
 )
 
 const (
@@ -343,12 +345,47 @@ func (cmvt *cdcMixedVersionTester) createChangeFeed(node int) versionStep {
 			{"updated", ""},
 			{"resolved", fmt.Sprintf("'%s'", resolvedInterval)},
 		}
-		_, err := newChangefeedCreator(db, fmt.Sprintf("%s.%s", targetDB, targetTable), cmvt.kafka.sinkURL(ctx)).
-			With(options...).
-			Create()
+		creator := newChangefeedCreator(db, fmt.Sprintf("%s.%s", targetDB, targetTable), cmvt.kafka.sinkURL(ctx)).
+			With(options...)
+
+		jobID, err := creator.Create()
 		if err != nil {
 			t.Fatal(err)
 		}
+		// In 22.1, a changefeed may fail during a rolling upgrade due to a bug
+		// which is fixed in 22.2 onwards. To accomodate for this bug,
+		// poll the changefeed job and restart it if it fails.
+		// See #106878 for more details.
+		go watchChangefeed(ctx, t, jobID, db, creator)
+	}
+}
+
+// watchChangefeed polls a changefeed job every 1 second and re-creates if it
+// failed.
+func watchChangefeed(
+	ctx context.Context, t test.Test, jobID int, db *gosql.DB, creator *changefeedCreator,
+) {
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-time.After(1 * time.Second):
+			var feedStatus string
+			if err := db.QueryRow(`SELECT status FROM system.jobs WHERE id = $1`, jobID).Scan(&feedStatus); err != nil {
+				t.Fatal(errors.Wrap(err, "changefeed watcher failed to query jobs table"))
+				return
+			}
+			if feedStatus == string(jobs.StatusFailed) {
+				t.L().Printf("changefeed %d failed. recreating changefeed", jobID)
+				newJobID, err := creator.Create()
+				if err != nil {
+					t.Fatal(errors.Wrap(err, "changefeed watcher failed to recreate changefeed job"))
+					return
+				}
+				jobID = newJobID
+				t.L().Printf("successfully recreated changefeed as new job %d", jobID)
+			}
+		}
 	}
 }