szetszwo commented on code in PR #9587:
URL: https://github.com/apache/ozone/pull/9587#discussion_r2662170250


##########
hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/diskbalancer/policy/DefaultVolumeChoosingPolicy.java:
##########
@@ -67,24 +67,41 @@ public Pair<HddsVolume, HddsVolume> 
chooseVolume(MutableVolumeSet volumeSet,
           .sorted(Comparator.comparingDouble(VolumeFixedUsage::getUtilization))
           .collect(Collectors.toList());
 
-      // Calculate the actual threshold and check src
-      final double actualThreshold = getIdealUsage(volumeUsages) + 
thresholdPercentage / 100;
-      final VolumeFixedUsage src = volumeUsages.get(volumeUsages.size() - 1);
-      if (src.getUtilization() < actualThreshold) {
-        return null; // all volumes are under the threshold
+      // Calculate ideal usage and threshold range
+      final double idealUsage = getIdealUsage(volumeUsages);
+      final double actualThreshold = thresholdPercentage / 100.0;
+      final double lowerThreshold = idealUsage - actualThreshold;
+      final double upperThreshold = idealUsage + actualThreshold;
+
+      // Get highest and lowest utilization volumes
+      final VolumeFixedUsage highestUsage = 
volumeUsages.get(volumeUsages.size() - 1);
+      final VolumeFixedUsage lowestUsage = volumeUsages.get(0);
+
+      // Only return null if highest is below upper threshold AND lowest is 
above lower threshold
+      // This means all volumes are strictly within the range (not at 
boundaries)
+      if (highestUsage.getUtilization() < upperThreshold && 
+          lowestUsage.getUtilization() > lowerThreshold) {
+        // All volumes are strictly within threshold range, no balancing needed
+        return null;
       }
 
-      // Find dst
+      // Determine source volume: highest utilization volume (if above 
threshold) 
+      final VolumeFixedUsage src = highestUsage;
+
+      // Find destination volume: lowest utilization volume that has enough 
space
+      // Prefer volumes below threshold, but accept any volume with lower 
utilization than source
       for (int i = 0; i < volumeUsages.size() - 1; i++) {

Review Comment:
   We should loop both src and dst.
   
   ```java
         for (int s = volumeUsages.size() - 1; s > 0; s--) {
           for (int d = 0; d < s; d++) {
             final VolumeFixedUsage srcUsage = volumeUsages.get(s);
             final VolumeFixedUsage dstUsage = volumeUsages.get(d);
             if (srcUsage.getUtilization() < upperThreshold && 
dstUsage.getUtilization() > lowerThreshold) {
               // within thresholds
               continue;
             }
   
             final HddsVolume dst = dstUsage.getVolume();
             if (containerSize < dstUsage.computeUsableSpace()) {
               // Found dst, reserve space and return
               dst.incCommittedBytes(containerSize);
               return Pair.of(srcUsage.getVolume(), dst);
             }
             LOG.debug("Destination volume {} does not have enough space, 
trying next volume.",
                 dst.getStorageID());
           }
         }
   ```



-- 
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]

Reply via email to