narendly commented on a change in pull request #639: Refine the WAGED
rebalancer to minimize the partial rebalance workload.
URL: https://github.com/apache/helix/pull/639#discussion_r352426262
##########
File path:
helix-core/src/main/java/org/apache/helix/controller/rebalancer/waged/model/ClusterModelProvider.java
##########
@@ -159,33 +337,29 @@ public static ClusterModel
generateClusterModelFromCurrentState(
.getOrDefault(HelixConstants.ChangeType.RESOURCE_CONFIG,
Collections.emptySet())
.contains(resourceName) || clusterChanges
.getOrDefault(HelixConstants.ChangeType.IDEAL_STATE,
Collections.emptySet())
- .contains(resourceName) ||
!bestPossibleAssignment.containsKey(resourceName)) {
+ .contains(resourceName) ||
!currentAssignment.containsKey(resourceName)) {
toBeAssignedReplicas.addAll(replicas);
continue; // go to check next resource
} else {
// check for every best possible assignments to identify if the
related replicas need to reassign.
- ResourceAssignment assignment =
bestPossibleAssignment.get(resourceName);
- // <partition, <instance, state>>
- Map<String, Map<String, String>> stateMap =
assignment.getMappedPartitions().stream()
- .collect(Collectors.toMap(partition ->
partition.getPartitionName(),
- partition -> new
HashMap<>(assignment.getReplicaMap(partition))));
+ // <partition, <state, instances list>>
+ Map<String, Map<String, Set<String>>> stateMap =
+ getValidStateInstanceMap(currentAssignment.get(resourceName),
activeInstances);
for (AssignableReplica replica : replicas) {
// Find any ACTIVE instance allocation that has the same state
with the replica
- Optional<Map.Entry<String, String>> instanceNameOptional =
- stateMap.getOrDefault(replica.getPartitionName(),
Collections.emptyMap()).entrySet()
- .stream().filter(instanceStateMap ->
-
instanceStateMap.getValue().equals(replica.getReplicaState()) && activeInstances
- .contains(instanceStateMap.getKey())).findAny();
- // 3. if no such an instance in the bestPossible assignment, need
to reassign the replica
- if (!instanceNameOptional.isPresent()) {
+ Set<String> validInstances =
+ stateMap.getOrDefault(replica.getPartitionName(),
Collections.emptyMap())
+ .getOrDefault(replica.getReplicaState(),
Collections.emptySet());
Review comment:
Chained `getOrDefault` statements are a pain to read/interpret. Can we just
do `contains()` where possible? I don't think it's a good idea to pass around
an `emptySet()` like it's a boolean, which you seem to be doing a lot in the
code. This applies not only to this piece of code, but PRs in general. Thanks :)
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
