[GitHub] [solr] noblepaul commented on a diff in pull request #1585: Use Zookeeper persistent recursive Zookeeper watches

["noblepaul (via GitHub)" <gi...@apache.org>]

noblepaul commented on code in PR #1585:
URL: https://github.com/apache/solr/pull/1585#discussion_r1191949973


##########
solr/solrj-zookeeper/src/java/org/apache/solr/common/cloud/ZkStateReader.java:
##########
@@ -368,10 +368,152 @@ private StatefulCollectionWatch compute(
         BiFunction<String, StatefulCollectionWatch, StatefulCollectionWatch> remappingFunction) {
       return statefulWatchesByCollectionName.compute(collectionName, remappingFunction);
     }
+
+    public boolean isWatched(String coll) {
+      return statefulWatchesByCollectionName.containsKey(coll);
+    }
   }
 
-  private static class StatefulCollectionWatch extends CollectionWatch<DocCollectionWatcher> {
+  private class StatefulCollectionWatch extends CollectionWatch<DocCollectionWatcher> {
     private DocCollection currentState;
+
+    Watcher persistentWatcher;
+
+    final String coll;
+    private final String collectionPath;
+
+    private StatefulCollectionWatch(String coll) {
+      this.coll = coll;
+      this.collectionPath = DocCollection.getCollectionPath(coll);
+    }
+
+    public void handleWatch(WatchedEvent event) {
+      if (!collectionWatches.isWatched(coll)) {
+        return;
+      }
+      if (log.isTraceEnabled()) {
+        log.trace("an event happened for {}, event: {}", coll, event.toString());
+      }
+      DocCollection collectionState = getCollection(coll);
+      if (collectionPath.equals(event.getPath())) {
+        DocCollection newState = null;
+        try {
+          newState = fetchCollectionState(coll, null);
+        } catch (KeeperException.SessionExpiredException
+            | KeeperException.ConnectionLossException e) {
+          log.warn("ZooKeeper watch triggered, but Solr cannot talk to ZK: ", e);
+        } catch (KeeperException e) {
+          log.error("exception for collection: [{}]", coll, e);
+          throw new ZooKeeperException(ErrorCode.SERVER_ERROR, "A ZK error has occurred", e);
+        } catch (InterruptedException e) {
+          Thread.currentThread().interrupt();
+          log.error("exception for collection: [{}]", coll, e);
+        }
+        collectionWatches.updateDocCollection(coll, newState);
+        synchronized (getUpdateLock()) {
+          constructState(Collections.singleton(coll));
+        }
+      }
+      if (collectionState != null && collectionState.isPerReplicaState()) {
+        String path = event.getPath();
+        if ((event.getType() == EventType.NodeCreated || event.getType() == EventType.NodeDeleted)
+            && path.length() > collectionState.getZNode().length()
+            && path.startsWith(collectionPath)) {
+          if (log.isDebugEnabled()) {
+            log.debug("PRS node event : {}", event.getType());
+          }
+
+          PerReplicaStates prs = collectionState.getPerReplicaStates();
+          if (prs != null) {
+            String stateStr = path.substring(collectionPath.length() + 1);
+            PerReplicaStates.State newState = PerReplicaStates.State.parse(stateStr);
+            PerReplicaStates.State oldState = prs.states.get(newState.replica);
+            if (event.getType() == EventType.NodeCreated) {
+              if (oldState != null && newState.version < oldState.version) {
+                // we got a notification out of order? . Shouldn't happen
+                if (log.isTraceEnabled()) {
+                  log.trace("newState {} < oldState {}", newState, oldState);
+                }
+                return;
+              }
+              if (oldState == null) {
+                // the state does not exist now. fetch everything
+
+                if (log.isTraceEnabled()) {
+                  log.trace("fresh replica, force fetch all {}", collectionPath);
+                }
+
+                prs = PerReplicaStatesFetcher.fetch(collectionPath, zkClient, null);
+              } else {
+                // the PRS entry is already available, need to do an update
+                Stat stat = null;
+                try {
+                  // get the stat of this child node
+                  stat = zkClient.exists(path, null, true);

Review Comment:
   that's incorrect . The stat of `state.json` would've gotten changed by another event



-- 
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: issues-unsubscribe@solr.apache.org

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


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@solr.apache.org
For additional commands, e-mail: issues-help@solr.apache.org