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

HDFS-16964. Improve processing of excess redundancy after failover. (#5510). Contributed by Shuyan Zhang.

Browse Source 
Signed-off-by: He Xiaoqiao <hexiaoqiao@apache.org>
This commit is contained in:
zhangshuyan 2023-03-28 16:14:59 +08:00 committed by GitHub
parent b5e8269d9b
commit 700147b4ac
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 38 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
View File

@ -3987,17 +3987,11 @@ private MisReplicationResult processMisReplicatedBlock(BlockInfo block) {
}
if (shouldProcessExtraRedundancy(num, expectedRedundancy)) {
if (num.replicasOnStaleNodes() > 0) {
// If any of the replicas of this block are on nodes that are
// considered "stale", then these replicas may in fact have
// already been deleted. So, we cannot safely act on the
// over-replication until a later point in time, when
// the "stale" nodes have block reported.
// extra redundancy block
if (!processExtraRedundancyBlockWithoutPostpone(block, expectedRedundancy,
null, null)) {
return MisReplicationResult.POSTPONE;
}
// extra redundancy block
processExtraRedundancyBlock(block, expectedRedundancy, null, null);
return MisReplicationResult.OVER_REPLICATED;
}
@ -4020,12 +4014,26 @@ public void setReplication(
}
}
/**
* Process blocks with redundant replicas. If there are replicas in
* stale storages, mark them in the postponedMisreplicatedBlocks.
*/
private void processExtraRedundancyBlock(final BlockInfo block,
final short replication, final DatanodeDescriptor addedNode,
DatanodeDescriptor delNodeHint) {
if (!processExtraRedundancyBlockWithoutPostpone(block, replication,
addedNode, delNodeHint)) {
postponeBlock(block);
}
}
/**
* Find how many of the containing nodes are "extra", if any.
* If there are any extras, call chooseExcessRedundancies() to
* mark them in the excessRedundancyMap.
* @return true if all redundancy replicas are removed.
*/
private void processExtraRedundancyBlock(final BlockInfo block,
private boolean processExtraRedundancyBlockWithoutPostpone(final BlockInfo block,
final short replication, final DatanodeDescriptor addedNode,
DatanodeDescriptor delNodeHint) {
assert namesystem.hasWriteLock();
@ -4035,17 +4043,17 @@ private void processExtraRedundancyBlock(final BlockInfo block,
Collection<DatanodeStorageInfo> nonExcess = new ArrayList<>();
Collection<DatanodeDescriptor> corruptNodes = corruptReplicas
.getNodes(block);
boolean hasStaleStorage = false;
Set<DatanodeStorageInfo> staleStorages = new HashSet<>();
for (DatanodeStorageInfo storage : blocksMap.getStorages(block)) {
if (storage.getState() != State.NORMAL) {
continue;
}
final DatanodeDescriptor cur = storage.getDatanodeDescriptor();
if (storage.areBlockContentsStale()) {
LOG.trace("BLOCK* processExtraRedundancyBlock: Postponing {}"
+ " since storage {} does not yet have up-to-date information.",
block, storage);
postponeBlock(block);
return;
hasStaleStorage = true;
staleStorages.add(storage);
continue;
}
if (!isExcess(cur, block)) {
if (cur.isInService()) {
@ -4058,6 +4066,13 @@ private void processExtraRedundancyBlock(final BlockInfo block,
}
chooseExcessRedundancies(nonExcess, block, replication, addedNode,
delNodeHint);
if (hasStaleStorage) {
LOG.trace("BLOCK* processExtraRedundancyBlockWithoutPostpone: Postponing {}"
+ " since storages {} does not yet have up-to-date information.",
block, staleStorages);
return false;
}
return true;
}
private void chooseExcessRedundancies(
@ -4071,12 +4086,14 @@ private void chooseExcessRedundancies(
if (storedBlock.isStriped()) {
chooseExcessRedundancyStriped(bc, nonExcess, storedBlock, delNodeHint);
} else {
final BlockStoragePolicy storagePolicy = storagePolicySuite.getPolicy(
bc.getStoragePolicyID());
final List<StorageType> excessTypes = storagePolicy.chooseExcess(
replication, DatanodeStorageInfo.toStorageTypes(nonExcess));
chooseExcessRedundancyContiguous(nonExcess, storedBlock, replication,
addedNode, delNodeHint, excessTypes);
if (nonExcess.size() > replication) {
final BlockStoragePolicy storagePolicy = storagePolicySuite.getPolicy(
bc.getStoragePolicyID());
final List<StorageType> excessTypes = storagePolicy.chooseExcess(
replication, DatanodeStorageInfo.toStorageTypes(nonExcess));
chooseExcessRedundancyContiguous(nonExcess, storedBlock, replication,
addedNode, delNodeHint, excessTypes);
}
}
}