horizonzy commented on code in PR #3359:
URL: https://github.com/apache/bookkeeper/pull/3359#discussion_r939807153
##########
bookkeeper-server/src/main/java/org/apache/bookkeeper/client/RackawareEnsemblePlacementPolicyImpl.java:
##########
@@ -1079,4 +1083,220 @@ public boolean
areAckedBookiesAdheringToPlacementPolicy(Set<BookieId> ackedBooki
}
return rackCounter.size() >= minWriteQuorumNumRacksPerWriteQuorum;
}
+
+ @Override
+ public PlacementResult<List<BookieId>> replaceToAdherePlacementPolicy(
+ int ensembleSize,
+ int writeQuorumSize,
+ int ackQuorumSize,
+ Set<BookieId> excludeBookies,
+ List<BookieId> currentEnsemble) {
+ rwLock.readLock().lock();
+ try {
+ PlacementPolicyAdherence currentPlacementAdherence =
isEnsembleAdheringToPlacementPolicy(
+ currentEnsemble, writeQuorumSize, ackQuorumSize);
+ if (PlacementPolicyAdherence.FAIL != currentPlacementAdherence) {
+ return PlacementResult.of(new ArrayList<>(currentEnsemble),
currentPlacementAdherence);
+ }
+ for (BookieId bookieId : currentEnsemble) {
+ if (!knownBookies.containsKey(bookieId)) {
+ excludeBookies.add(bookieId);
+ }
+ }
+ PlacementResult<List<BookieId>> placementResult =
PlacementResult.of(Collections.emptyList(),
+ PlacementPolicyAdherence.FAIL);
+ int minDiffer = Integer.MAX_VALUE;
+ for (int i = 0; i < currentEnsemble.size(); i++) {
+ PlacementResult<List<BookieId>> result =
doReplaceToAdherePlacementPolicy(ensembleSize,
+ writeQuorumSize, ackQuorumSize, excludeBookies,
currentEnsemble, i);
+ if (PlacementPolicyAdherence.FAIL ==
result.getAdheringToPolicy()) {
+ continue;
+ }
+ int differ = differBetweenBookies(currentEnsemble,
result.getResult());
+ if (differ < minDiffer) {
+ minDiffer = differ;
+ placementResult = result;
+ if (minDiffer == 1) {
+ break;
+ }
+ }
+ }
+ return placementResult;
+ } finally {
+ rwLock.readLock().unlock();
+ }
+ }
+
+ private PlacementResult<List<BookieId>> doReplaceToAdherePlacementPolicy(
+ int ensembleSize,
+ int writeQuorumSize,
+ int ackQuorumSize,
+ Set<BookieId> excludeBookies,
+ List<BookieId> currentEnsemble,
+ int startIndex) {
+ final List<BookieNode> provisionalEnsembleNodes =
currentEnsemble.stream()
+ .map(this::convertBookieToNode).collect(Collectors.toList());
+ final Set<Node> excludeNodes = convertBookiesToNodes(
+ addDefaultRackBookiesIfMinNumRacksIsEnforced(excludeBookies));
+ int minNumRacksPerWriteQuorumForThisEnsemble =
Math.min(writeQuorumSize, minNumRacksPerWriteQuorum);
+ final RRTopologyAwareCoverageEnsemble ensemble =
+ new RRTopologyAwareCoverageEnsemble(
+ ensembleSize,
+ writeQuorumSize,
+ ackQuorumSize,
+ RACKNAME_DISTANCE_FROM_LEAVES,
+ null,
+ null,
+ minNumRacksPerWriteQuorumForThisEnsemble);
+ int numRacks = topology.getNumOfRacks();
+ // only one rack or less than
minNumRacksPerWriteQuorumForThisEnsemble, stop calculation to skip relocation
+ if (numRacks < 2 || numRacks <
minNumRacksPerWriteQuorumForThisEnsemble) {
+ LOG.warn("Skip ensemble relocation because the cluster has only {}
rack.", numRacks);
+ return PlacementResult.of(Collections.emptyList(),
PlacementPolicyAdherence.FAIL);
+ }
+ BookieNode prevNode = null;
+ final BookieNode firstNode = provisionalEnsembleNodes.get(startIndex);
+ // use same bookie at first to reduce ledger replication
+ if (!excludeNodes.contains(firstNode) && ensemble.apply(firstNode,
ensemble)
+ && ensemble.addNode(firstNode)) {
+ excludeNodes.add(firstNode);
+ prevNode = firstNode;
+ }
Review Comment:
Do you means `provisionalEnsembleNodes.set(0, firstNode)`?
If so, could you explain it more? We just change the start index to
calculate, I think we shouldn't change `provisionalEnsembleNodes` element
##########
bookkeeper-server/src/main/java/org/apache/bookkeeper/net/NetworkTopology.java:
##########
@@ -29,6 +29,7 @@ public interface NetworkTopology {
String DEFAULT_RACK = "/default-rack";
String DEFAULT_ZONE = "/default-zone";
String DEFAULT_UPGRADEDOMAIN = "/default-upgradedomain";
+ String INACTIVE = "/inactive";
Review Comment:
nice catch
##########
bookkeeper-server/src/main/java/org/apache/bookkeeper/client/RackawareEnsemblePlacementPolicyImpl.java:
##########
@@ -1079,4 +1083,220 @@ public boolean
areAckedBookiesAdheringToPlacementPolicy(Set<BookieId> ackedBooki
}
return rackCounter.size() >= minWriteQuorumNumRacksPerWriteQuorum;
}
+
+ @Override
+ public PlacementResult<List<BookieId>> replaceToAdherePlacementPolicy(
+ int ensembleSize,
+ int writeQuorumSize,
+ int ackQuorumSize,
+ Set<BookieId> excludeBookies,
+ List<BookieId> currentEnsemble) {
+ rwLock.readLock().lock();
+ try {
+ PlacementPolicyAdherence currentPlacementAdherence =
isEnsembleAdheringToPlacementPolicy(
+ currentEnsemble, writeQuorumSize, ackQuorumSize);
+ if (PlacementPolicyAdherence.FAIL != currentPlacementAdherence) {
+ return PlacementResult.of(new ArrayList<>(currentEnsemble),
currentPlacementAdherence);
+ }
+ for (BookieId bookieId : currentEnsemble) {
+ if (!knownBookies.containsKey(bookieId)) {
+ excludeBookies.add(bookieId);
+ }
+ }
+ PlacementResult<List<BookieId>> placementResult =
PlacementResult.of(Collections.emptyList(),
+ PlacementPolicyAdherence.FAIL);
+ int minDiffer = Integer.MAX_VALUE;
+ for (int i = 0; i < currentEnsemble.size(); i++) {
+ PlacementResult<List<BookieId>> result =
doReplaceToAdherePlacementPolicy(ensembleSize,
+ writeQuorumSize, ackQuorumSize, excludeBookies,
currentEnsemble, i);
+ if (PlacementPolicyAdherence.FAIL ==
result.getAdheringToPolicy()) {
+ continue;
+ }
+ int differ = differBetweenBookies(currentEnsemble,
result.getResult());
+ if (differ < minDiffer) {
+ minDiffer = differ;
+ placementResult = result;
+ if (minDiffer == 1) {
+ break;
+ }
+ }
+ }
+ return placementResult;
+ } finally {
+ rwLock.readLock().unlock();
+ }
+ }
+
+ private PlacementResult<List<BookieId>> doReplaceToAdherePlacementPolicy(
+ int ensembleSize,
+ int writeQuorumSize,
+ int ackQuorumSize,
+ Set<BookieId> excludeBookies,
+ List<BookieId> currentEnsemble,
+ int startIndex) {
+ final List<BookieNode> provisionalEnsembleNodes =
currentEnsemble.stream()
+ .map(this::convertBookieToNode).collect(Collectors.toList());
+ final Set<Node> excludeNodes = convertBookiesToNodes(
+ addDefaultRackBookiesIfMinNumRacksIsEnforced(excludeBookies));
+ int minNumRacksPerWriteQuorumForThisEnsemble =
Math.min(writeQuorumSize, minNumRacksPerWriteQuorum);
+ final RRTopologyAwareCoverageEnsemble ensemble =
+ new RRTopologyAwareCoverageEnsemble(
+ ensembleSize,
+ writeQuorumSize,
+ ackQuorumSize,
+ RACKNAME_DISTANCE_FROM_LEAVES,
+ null,
+ null,
+ minNumRacksPerWriteQuorumForThisEnsemble);
+ int numRacks = topology.getNumOfRacks();
+ // only one rack or less than
minNumRacksPerWriteQuorumForThisEnsemble, stop calculation to skip relocation
+ if (numRacks < 2 || numRacks <
minNumRacksPerWriteQuorumForThisEnsemble) {
+ LOG.warn("Skip ensemble relocation because the cluster has only {}
rack.", numRacks);
+ return PlacementResult.of(Collections.emptyList(),
PlacementPolicyAdherence.FAIL);
+ }
Review Comment:
yes
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
