xl: Add stat calls to keep track of ignored errors. (#4117)

Such that in a situation where all errors were
ignored we need to reduce the errors using
readQuorum to get a consistent error value.

Without this change errors generated will
never be consistent with for an expected scenario.

For example in a 6 disk setup 1 disk is missing
and 5 do not have the volume (testbucket)

Without this change Stat() would result in different
errors depending on which disk died. Can cause
confusion to S3 client application.

This change addresses need to track type of
errors we ignored and bring readQuorum to
choose the maximally occuring as the value
of truth.

cmd/xl-v1-multipart.go

@@ -208,25 +208,28 @@ func (xl xlObjects) removeObjectPart(bucket, object, uploadID, partName string)
 
 // statPart - returns fileInfo structure for a successful stat on part file.
 func (xl xlObjects) statPart(bucket, object, uploadID, partName string) (fileInfo FileInfo, err error) {
+	var ignoredErrs []error
 	partNamePath := path.Join(bucket, object, uploadID, partName)
 	for _, disk := range xl.getLoadBalancedDisks() {
 		if disk == nil {
+			ignoredErrs = append(ignoredErrs, errDiskNotFound)
 			continue
 		}
 		fileInfo, err = disk.StatFile(minioMetaMultipartBucket, partNamePath)
 		if err == nil {
 			return fileInfo, nil
 		}
-		err = traceError(err)
 		// For any reason disk was deleted or goes offline we continue to next disk.
 		if isErrIgnored(err, objMetadataOpIgnoredErrs...) {
+			ignoredErrs = append(ignoredErrs, err)
 			continue
 		}
-
-		// Catastrophic error, we return.
-		break
+		// Error is not ignored, return right here.
+		return FileInfo{}, traceError(err)
 	}
-	return FileInfo{}, err
+	// If all errors were ignored, reduce to maximal occurrence
+	// based on the read quorum.
+	return FileInfo{}, reduceReadQuorumErrs(ignoredErrs, nil, xl.readQuorum)
 }
 
 // commitXLMetadata - commit `xl.json` from source prefix to destination prefix in the given slice of disks.