[
https://issues.apache.org/jira/browse/ZOOKEEPER-1458?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Patrick Hunt resolved ZOOKEEPER-1458.
-------------------------------------
Resolution: Not A Problem
Assignee: Patrick Hunt
Yes, this is expected in 3.4+ (cversion/assignment behavior changed slightly to
fix a bug). See the discussion on this jira, esp starting here:
https://issues.apache.org/jira/browse/ZOOKEEPER-1046?focusedCommentId=13049607&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13049607
also this comment:
https://issues.apache.org/jira/browse/ZOOKEEPER-1046?focusedCommentId=13050222&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13050222
> Parent's cversion doesn't match the sequence number that get assigned to a
> child node with the SEQUENTIAL flag on.
> ------------------------------------------------------------------------------------------------------------------
>
> Key: ZOOKEEPER-1458
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1458
> Project: ZooKeeper
> Issue Type: Bug
> Components: server
> Affects Versions: 3.4.3
> Environment: All
> Reporter: Andrey Kornev
> Assignee: Patrick Hunt
>
> If I have a child delete op interleaving two child create ops, the second
> child create will nevertheless have the path suffix incremented only by 1
> rather than by 2. Is this expected? The 3.3.5 version takes into account the
> delete and increments the sequence by 2.
> PrepRequestProcessor uses the parent's cversion to generate the child's
> sequence suffix. However it appears that this particular cversion only counts
> "create" operations and it doesn't take into account the deletes. Strangely
> enough, the parent stats returned by getData() show the correct cversion with
> all the creates and deletes accounted for.
> It looks like the first cversion comes from the ChangeRecord for the parent
> node stuck in ZooKeeperServer.outstandingChangesForPath map. And the second
> one (returned by getData(), that is) comes from the DataTree.
> Here's a simple example that reproduces the situation.
> zk.create("/parent", null, OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
> Stat stat = new Stat();
> zk.getData("/parent", false, stat);
> stat.getCVersion(); // returns 0 -- expected;
> String actualPath = zk.create("/parent/child", null, OPEN_ACL_UNSAFE,
> CreateMode.EPHEMERAL_SEQUENTIAL);
> // actualPath is "/parent/child0000000000" -- expected.
> zk.getData("/parent", false, stat);
> stat.getCVersion(); // returns 1 -- expected;
> zk.getData(actualPath, false, stat);
> zk.delete(actualPath,stat.getVersion()); // delete the child node
> zk.getData("/parent", false, stat);
> stat.getCVersion(); // returns 2;
> // create another child
> actualPath = zk.create("/parent/child", null, OPEN_ACL_UNSAFE,
> CreateMode.EPHEMERAL_SEQUENTIAL);
> // returned "/parent/child0000000001" but expected "/parent/child0000000002"
> zk.getData("/parent", false, stat);
> stat.getCVersion(); // returns 3;
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
