somandal commented on code in PR #15266:
URL: https://github.com/apache/pinot/pull/15266#discussion_r2029524839


##########
pinot-controller/src/main/java/org/apache/pinot/controller/helix/core/rebalance/TableRebalanceProgressStats.java:
##########
@@ -118,6 +128,165 @@ public RebalanceStateStats 
getCurrentToTargetConvergence() {
     return _currentToTargetConvergence;
   }
 
+  public RebalanceProgressStats getRebalanceProgressStatsOverall() {
+    return _rebalanceProgressStatsOverall;
+  }
+
+  public RebalanceProgressStats getRebalanceProgressStatsCurrentStep() {
+    return _rebalanceProgressStatsCurrentStep;
+  }
+
+  /**
+   * Updates the overall and step progress stats based on the latest 
calculated step's progress stats. This should
+   * be called during the EV-IS convergence trigger to ensure the overall 
stats reflect the changes as they are made.
+   * @param currentStepStats latest step level stats calculated in this 
iteration
+   */
+  public void updateOverallAndStepStatsFromLatestStepStats(
+      TableRebalanceProgressStats.RebalanceProgressStats currentStepStats) {
+    // Fetch the step level and overall stats that were calculated in the last 
convergence check. These will be used
+    // to calculate the overall stats in the current convergence check
+    TableRebalanceProgressStats.RebalanceProgressStats previousStepStats = 
getRebalanceProgressStatsCurrentStep();
+    TableRebalanceProgressStats.RebalanceProgressStats previousOverallStats = 
getRebalanceProgressStatsOverall();
+
+    // Calculate the new segments added / deleted since the last step to track 
overall change in segments. These are
+    // only new segments tracked by table rebalance (segments that aren't 
tracked by rebalance convergence check are
+    // ignored because the convergence check does not wait for them to 
complete)
+    int numAdditionalSegmentsAdded =
+        currentStepStats._totalSegmentsToBeAdded - 
previousStepStats._totalSegmentsToBeAdded;
+    int numAdditionalSegmentsDeleted =
+        currentStepStats._totalSegmentsToBeDeleted - 
previousStepStats._totalSegmentsToBeDeleted;
+
+    // Calculate the carry-over segment adds / deletes from the previous 
rebalance step if any (can happen if
+    // bestEfforts=true or if there are deletions left over from the last 
step) and the difference in the number of
+    // segments processed since the previous stats update.
+    // Example of what carry-over will look like:
+    //     previousStepStats: totalSegmentsToBeAdded = 10
+    //     previousStepStats: totalRemainingSegmentsToBeAdded = 15 (this can 
only be larger than totalSegmentsToBeAdded
+    //                                                              if 
carry-over segments exist)
+    //     Carry-over: 15 - 10 = 5 segments -> which were added in the last IS 
update by rebalance, but which didn't
+    //                 finish getting added before making the next assignment 
IS update
+    // Using the above, we calculate the 'upperBoundOnSegmentsAdded/Deleted' 
as the maximum segments to use when
+    // calculating the number of segments processed (number of segments that 
got added / deleted in EV based on expected
+    // IS) since the last step. We want to skip including the carry-over 
segments if any as they are tracked separately.
+    int upperBoundOnSegmentsAdded =
+        previousStepStats._totalRemainingSegmentsToBeAdded > 
previousStepStats._totalSegmentsToBeAdded
+            ? previousStepStats._totalSegmentsToBeAdded : 
previousStepStats._totalRemainingSegmentsToBeAdded;
+    int upperBoundOnSegmentsDeleted =
+        previousStepStats._totalRemainingSegmentsToBeDeleted > 
previousStepStats._totalSegmentsToBeDeleted
+            ? previousStepStats._totalSegmentsToBeDeleted : 
previousStepStats._totalRemainingSegmentsToBeDeleted;
+
+    // Number of adds / deletes processed in the last step. Take the absolute 
here since if new segments were added /
+    // deleted in the current step that are being tracking, we need to treat 
those as new segments that remain to
+    // be processed.
+    int numSegmentAddsProcessedInLastStep = Math.abs(upperBoundOnSegmentsAdded
+        - currentStepStats._totalRemainingSegmentsToBeAdded);
+    int numSegmentDeletesProcessedInLastStep = 
Math.abs(upperBoundOnSegmentsDeleted
+        - currentStepStats._totalRemainingSegmentsToBeDeleted);
+
+    // Number of untracked segments that were added
+    int numberNewUntrackedSegmentsAdded = 
currentStepStats._totalUniqueNewUntrackedSegmentsDuringRebalance
+        - previousStepStats._totalUniqueNewUntrackedSegmentsDuringRebalance;
+
+    TableRebalanceProgressStats.RebalanceProgressStats currentOverallStats =
+        new TableRebalanceProgressStats.RebalanceProgressStats();
+
+    // New number of total segment adds / deletes should include any new 
segments added from the previous stats
+    currentOverallStats._totalSegmentsToBeAdded = 
previousOverallStats._totalSegmentsToBeAdded
+        + numAdditionalSegmentsAdded;
+    currentOverallStats._totalSegmentsToBeDeleted = 
previousOverallStats._totalSegmentsToBeDeleted
+        + numAdditionalSegmentsDeleted;
+
+    // If carry-over segments are present, keep the upper bound of remaining 
segments to be added / deleted at the same
+    // level as the previous stats. This is so we track progress to handle the 
carry-over segments before we start
+    // tracking progress against the segments that need to be handled as part 
of the current rebalance step.
+    if (currentStepStats._totalCarryOverSegmentsToBeAdded > 0) {
+      currentOverallStats._totalRemainingSegmentsToBeAdded = 
previousOverallStats._totalRemainingSegmentsToBeAdded;
+    } else {
+      // If additional segments were added in the previous step, these need to 
be added to the remaining to be added
+      // number as they are new segments that should be tracked for completion
+      // Example:
+      //     currentStepStats: totalSegmentsToBeAdded = 20
+      //     previousStepStats: totalSegmentsToBeAdded = 10
+      //     numAdditionalSegmentsAdded = 20 - 10 = 10
+      // In the above scenario, the new 'numAdditionalSegmentsAdded' also need 
to be tracked as
+      // 'totalRemainingSegmentsToBeAdded' for the current step, thus we need 
to add the number of segments processed
+      // If 'numAdditionalSegmentsAdded' = 0, that means no new segments were 
added in the previous step, and if there
+      // was a change in the segments processed in the last step, they 
completed processing
+      currentOverallStats._totalRemainingSegmentsToBeAdded = 
numAdditionalSegmentsAdded == 0
+          ? previousOverallStats._totalRemainingSegmentsToBeAdded - 
numSegmentAddsProcessedInLastStep
+          : previousOverallStats._totalRemainingSegmentsToBeAdded + 
numSegmentAddsProcessedInLastStep;

Review Comment:
   good catch, updated, please take a look



-- 
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: commits-unsubscr...@pinot.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org
For additional commands, e-mail: commits-h...@pinot.apache.org

Reply via email to