lei-xia commented on a change in pull request #1413:
URL: https://github.com/apache/helix/pull/1413#discussion_r499748505



##########
File path: 
helix-core/src/main/java/org/apache/helix/controller/dataproviders/BaseControllerDataProvider.java
##########
@@ -299,6 +306,33 @@ private void updateMaintenanceInfo(final HelixDataAccessor 
accessor) {
     // The following flag is to guarantee that there's only one update per 
pineline run because we
     // check for whether maintenance recovery could happen twice every pipeline
     _hasMaintenanceSignalChanged = false;
+
+    // If maintenance mode has exited, clear cached timed-out nodes
+    if (!_isMaintenanceModeEnabled) {
+      _timedOutInstanceDuringMaintenance.clear();
+    }
+  }
+
+  private void timeoutNodesDuringMaintenance(final HelixDataAccessor accessor) 
{
+    // If maintenance mode is enabled and timeout window is specified, filter 
'new' live nodes
+    // for timed-out nodes
+    long timeOutWindow = -1;
+    if (_clusterConfig != null) {
+      timeOutWindow = _clusterConfig.getOfflineNodeTimeOutForMaintenanceMode();
+    }
+    if (timeOutWindow >= 0 && isMaintenanceModeEnabled()) {
+      for (String instance : _liveInstanceCache.getPropertyMap().keySet()) {
+        // 1. Check timed-out cache and don't do repeated work;
+        // 2. Check for nodes that didn't exist in the last iteration, because 
it has been checked;
+        // 3. For all other nodes, check if it's timed-out.
+        // When maintenance mode is first entered, all nodes will be checked 
as a result.
+        if (!_timedOutInstanceDuringMaintenance.contains(instance)
+            && !_liveInstanceSnapshotForMaintenance.containsKey(instance)
+            && isInstanceTimedOutDuringMaintenance(accessor, instance, 
timeOutWindow)) {
+          _timedOutInstanceDuringMaintenance.add(instance);

Review comment:
       There could be some issue with this if a long-offline instance (say 
InstanceA) came back just after the cluster enter maintenance mode and before 
the next pipeline started.  InstanceA will be in the 
_liveInstanceSnapshotForMaintenance set (see line 261), so it will skip the 
timeout check here. 




----------------------------------------------------------------
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]



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to