NealSun96 commented on code in PR #2289:
URL: https://github.com/apache/helix/pull/2289#discussion_r1029681292
##########
helix-core/src/main/java/org/apache/helix/controller/rebalancer/waged/AssignmentMetadataStore.java:
##########
@@ -61,36 +62,42 @@ protected AssignmentMetadataStore(BucketDataAccessor
bucketDataAccessor, String
_bestPossiblePath = String.format(BEST_POSSIBLE_TEMPLATE, clusterName,
ASSIGNMENT_METADATA_KEY);
}
- public synchronized Map<String, ResourceAssignment> getBaseline() {
+ public Map<String, ResourceAssignment> getBaseline() {
// Return the in-memory baseline. If null, read from ZK. This is to
minimize reads from ZK
if (_globalBaseline == null) {
- try {
- HelixProperty baseline =
- _dataAccessor.compressedBucketRead(_baselinePath,
HelixProperty.class);
- _globalBaseline = splitAssignments(baseline);
- } catch (ZkNoNodeException ex) {
- // Metadata does not exist, so return an empty map
- _globalBaseline = new HashMap<>();
+ // double-checked locking
+ synchronized (this) {
+ if (_globalBaseline == null) {
+ _globalBaseline = fetchAssignmentOrDefault(_baselinePath);
+ }
}
}
return _globalBaseline;
}
- public synchronized Map<String, ResourceAssignment>
getBestPossibleAssignment() {
+ public Map<String, ResourceAssignment> getBestPossibleAssignment() {
// Return the in-memory baseline. If null, read from ZK. This is to
minimize reads from ZK
if (_bestPossibleAssignment == null) {
- try {
- HelixProperty baseline =
- _dataAccessor.compressedBucketRead(_bestPossiblePath,
HelixProperty.class);
- _bestPossibleAssignment = splitAssignments(baseline);
- } catch (ZkNoNodeException ex) {
- // Metadata does not exist, so return an empty map
- _bestPossibleAssignment = new HashMap<>();
+ // double-checked locking
+ synchronized (this) {
+ if (_bestPossibleAssignment == null) {
+ _bestPossibleAssignment =
fetchAssignmentOrDefault(_bestPossiblePath);
+ }
}
}
return _bestPossibleAssignment;
}
+ private Map<String, ResourceAssignment> fetchAssignmentOrDefault(String
path) {
+ try {
Review Comment:
My reasoning was that this function was synchronized before your change (and
even your change wraps this in a synchronized block), so it should continue to
be synchronized down the road.
You did raise a good point - my impression is bucket operations are not
thread safe, but maybe we want to verify that.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
