HDFS-15644. Failed volumes can cause DNs to stop block reporting. Contributed by Ahmed Hussein.

Author: jojochuang

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java

@@ -1957,28 +1957,32 @@ public Map<DatanodeStorage, BlockListAsLongs> getBlockReports(String bpid) {
           continue;
         }
         String volStorageID = b.getVolume().getStorageID();
-        if (!builders.containsKey(volStorageID)) {
-          if (!missingVolumesReported.contains(volStorageID)) {
-            LOG.warn("Storage volume: " + volStorageID + " missing for the"
-                + " replica block: " + b + ". Probably being removed!");
-            missingVolumesReported.add(volStorageID);
-          }
-          continue;
-        }
         switch(b.getState()) {
         case FINALIZED:
         case RBW:
         case RWR:
-          builders.get(volStorageID).add(b);
           break;
         case RUR:
-          ReplicaInfo orig = b.getOriginalReplica();
-          builders.get(volStorageID).add(orig);
+          // use the original replica.
+          b = b.getOriginalReplica();
           break;
         case TEMPORARY:
-          break;
+          continue;
         default:
           assert false : "Illegal ReplicaInfo state.";
+          continue;
+        }
+        BlockListAsLongs.Builder storageBuilder = builders.get(volStorageID);
+        // a storage in the process of failing will not be in the volumes list
+        // but will be in the replica map.
+        if (storageBuilder != null) {
+          storageBuilder.add(b);
+        } else {
+          if (!missingVolumesReported.contains(volStorageID)) {
+            LOG.warn("Storage volume: " + volStorageID + " missing for the"
+                + " replica block: " + b + ". Probably being removed!");
+            missingVolumesReported.add(volStorageID);
+          }
         }
       }
     }