big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

HDFS-10477. Stop decommission a rack of DataNodes caused NameNode fail over to standby. Contributed by yunjiong zhao and Wei-Chiu Chuang.

Browse Source
This commit is contained in:
Wei-Chiu Chuang 2019-04-03 11:00:12 -07:00
parent 3b0c5016b2
commit be488b6070
1 changed files with 32 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
View File

@ -4235,21 +4235,41 @@ void processExtraRedundancyBlocksOnInService(
if (!isPopulatingReplQueues()) { if (!isPopulatingReplQueues()) {
return; return;
} }
final Iterator<BlockInfo> it = srcNode.getBlockIterator();
int numExtraRedundancy = 0; int numExtraRedundancy = 0;
while(it.hasNext()) { for (DatanodeStorageInfo datanodeStorageInfo : srcNode.getStorageInfos()) {
final BlockInfo block = it.next(); // the namesystem lock is released between iterations. Make sure the
if (block.isDeleted()) { // storage is not removed before continuing.
//Orphan block, will be handled eventually, skip if (srcNode.getStorageInfo(datanodeStorageInfo.getStorageID()) == null) {
continue; continue;
} }
int expectedReplication = this.getExpectedRedundancyNum(block); final Iterator<BlockInfo> it = datanodeStorageInfo.getBlockIterator();
NumberReplicas num = countNodes(block); while(it.hasNext()) {
if (shouldProcessExtraRedundancy(num, expectedReplication)) { final BlockInfo block = it.next();
// extra redundancy block if (block.isDeleted()) {
processExtraRedundancyBlock(block, (short) expectedReplication, null, //Orphan block, will be handled eventually, skip
null); continue;
numExtraRedundancy++; }
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 " LOG.info("Invalidated {} extra redundancy blocks on {} after "