From be488b6070a124234c77f16193ee925d32ca9a20 Mon Sep 17 00:00:00 2001
From: Wei-Chiu Chuang <weichiu@apache.org>
Date: Wed, 3 Apr 2019 11:00:12 -0700
Subject: [PATCH] HDFS-10477. Stop decommission a rack of DataNodes caused
 NameNode fail over to standby. Contributed by  yunjiong zhao and Wei-Chiu
 Chuang.

---
 .../server/blockmanagement/BlockManager.java  | 44 ++++++++++++++-----
 1 file changed, 32 insertions(+), 12 deletions(-)

diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
index 4d3f1d1a51..4e351c021b 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
@@ -4235,21 +4235,41 @@ void processExtraRedundancyBlocksOnInService(
     if (!isPopulatingReplQueues()) {
       return;
     }
-    final Iterator<BlockInfo> it = srcNode.getBlockIterator();
+
     int numExtraRedundancy = 0;
-    while(it.hasNext()) {
-      final BlockInfo block = it.next();
-      if (block.isDeleted()) {
-        //Orphan block, will be handled eventually, skip
+    for (DatanodeStorageInfo datanodeStorageInfo : srcNode.getStorageInfos()) {
+      // the namesystem lock is released between iterations. Make sure the
+      // storage is not removed before continuing.
+      if (srcNode.getStorageInfo(datanodeStorageInfo.getStorageID()) == null) {
         continue;
       }
-      int expectedReplication = this.getExpectedRedundancyNum(block);
-      NumberReplicas num = countNodes(block);
-      if (shouldProcessExtraRedundancy(num, expectedReplication)) {
-        // extra redundancy block
-        processExtraRedundancyBlock(block, (short) expectedReplication, null,
-            null);
-        numExtraRedundancy++;
+      final Iterator<BlockInfo> it = datanodeStorageInfo.getBlockIterator();
+      while(it.hasNext()) {
+        final BlockInfo block = it.next();
+        if (block.isDeleted()) {
+          //Orphan block, will be handled eventually, skip
+          continue;
+        }
+        int expectedReplication = this.getExpectedRedundancyNum(block);
+        NumberReplicas num = countNodes(block);
+        if (shouldProcessExtraRedundancy(num, expectedReplication)) {
+          // extra redundancy block
+          processExtraRedundancyBlock(block, (short) expectedReplication, null,
+              null);
+          numExtraRedundancy++;
+        }
+      }
+      // When called by tests like TestDefaultBlockPlacementPolicy.
+      // testPlacementWithLocalRackNodesDecommissioned, it is not protected by
+      // lock, only when called by DatanodeManager.refreshNodes have writeLock
+      if (namesystem.hasWriteLock()) {
+        namesystem.writeUnlock();
+        try {
+          Thread.sleep(1);
+        } catch (InterruptedException e) {
+          Thread.currentThread().interrupt();
+        }
+        namesystem.writeLock();
       }
     }
     LOG.info("Invalidated {} extra redundancy blocks on {} after "