skip prevoting for decommissioned nodes irrespective of cluster state

imRishN · Sep 1, 2022 · 5602013 · 5602013
1 parent 8eb27f0
commit 5602013
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 1 deletion.
diff --git a/server/src/main/java/org/opensearch/cluster/coordination/Coordinator.java b/server/src/main/java/org/opensearch/cluster/coordination/Coordinator.java
@@ -1439,6 +1439,11 @@ private void startElectionScheduler() {
             public void run() {
                 synchronized (mutex) {
                     if (mode == Mode.CANDIDATE) {
+                        if(peerFinder.localNodeDecommissioned()) {
+                            logger.debug("skip prevoting as local node is decommissioned");
+                            return;
+                        }
+
                         final ClusterState lastAcceptedState = coordinationState.get().getLastAcceptedState();
 
                         if (localNodeMayWinElection(lastAcceptedState) == false) {

diff --git a/server/src/main/java/org/opensearch/discovery/PeerFinder.java b/server/src/main/java/org/opensearch/discovery/PeerFinder.java
@@ -110,6 +110,7 @@ public abstract class PeerFinder {
 
     private volatile long currentTerm;
     private boolean active;
+    private boolean localNodeDecommissioned = false;
     private DiscoveryNodes lastAcceptedNodes;
     private final Map<TransportAddress, Peer> peersByAddress = new LinkedHashMap<>();
     private Optional<DiscoveryNode> leader = Optional.empty();
@@ -143,17 +144,27 @@ public ActionListener<Void> nodeCommissionedListener() {
             @Override
             public void onResponse(Void unused) {
                 logger.info("setting findPeersInterval to [{}], due to recommissioning", findPeersInterval);
+                assert localNodeDecommissioned; // TODO: Do we need this?
+                localNodeDecommissioned = false;
                 findPeersInterval = DISCOVERY_FIND_PEERS_INTERVAL_SETTING.get(settings);
+
             }
 
             @Override
             public void onFailure(Exception e) {
-                logger.info("setting findPeersInterval to [{}], due to decommissioning", findPeersInterval);
+                logger.info("setting findPeersInterval to [{}], due to decommissioning",
+                    DISCOVERY_FIND_PEERS_INTERVAL_DURING_DECOMMISSION_SETTING.get(settings));
+                assert !localNodeDecommissioned;
+                localNodeDecommissioned = true;
                 findPeersInterval = DISCOVERY_FIND_PEERS_INTERVAL_DURING_DECOMMISSION_SETTING.get(settings);
             }
         };
     }
 
+    public boolean localNodeDecommissioned() {
+        return localNodeDecommissioned;
+    }
+
     public void activate(final DiscoveryNodes lastAcceptedNodes) {
         logger.trace("activating with {}", lastAcceptedNodes);