[
https://issues.apache.org/jira/browse/ZOOKEEPER-645?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12805736#action_12805736
]
Robert Crocombe commented on ZOOKEEPER-645:
-------------------------------------------
I can address point (1) in the original comment: having the session ID breaks
the lock recipe. It looks like the session ID was added per:
https://issues.apache.org/jira/browse/ZOOKEEPER-78?focusedCommentId=12615305&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#action_12615305
but the sorting via ZNodeName is worse than starvation: it causes clients with
low session IDs not to see that clients with higher session IDs have acquired
the lock which causes the function to return true: so you could theoretically
have as many lock owners as clients as long as they acquire the lock in
descending session ID order.
I fixed this in a local branch by replacing ZNodeName with a different class
that ignores the session ID when sorting. My test for WriteLockTest became:
{noformat}
/*
* Test that the bug which makes it possible to acquire a lock in two
* different sessions simultaneously is fixed. Bug occurs because
including
* session ID in node name results in sorting in
* LockZooKeeperOperation.execute() such that low session ID clients do
not
* see that higher session ID clients already have the lock.
*/
@Test
public void testSessionOrderingBugFix() throws Exception {
// session IDs are presumably assigned in order, but perform
some checks
final ZooKeeper zooA = createClient();
final ZooKeeper zooB = createClient();
final ZooKeeper lowZoo = zooA.getSessionId() <
zooB.getSessionId() ? zooA
: zooB;
final ZooKeeper highZoo = zooA.getSessionId() <
zooB.getSessionId() ? zooB
: zooA;
final WriteLock highLock = new WriteLock(highZoo, dir, null);
final WriteLock lowLock = new WriteLock(lowZoo, dir, null);
boolean highLocked = highLock.lock();
assertTrue(highLocked);
// If the bug is present, this attempt to lock will succeed.
boolean lowLocked = lowLock.lock();
assertFalse(lowLocked);
assertTrue(highLock.isOwner());
assertFalse(lowLock.isOwner());
}
{noformat}
> Bug in WriteLock recipe implementation?
> ---------------------------------------
>
> Key: ZOOKEEPER-645
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-645
> Project: Zookeeper
> Issue Type: Bug
> Components: recipes
> Affects Versions: 3.2.2
> Environment: 3.2.2 java 1.6.0_12
> Reporter: Jaakko Laine
> Assignee: Jaakko Laine
> Priority: Minor
> Fix For: 3.3.0
>
> Attachments: 645-fix-findPrefixInChildren.patch
>
>
> Not sure, but there seem to be two issues in the example WriteLock:
> (1) ZNodeName is sorted according to session ID first, and then according to
> znode sequence number. This might cause starvation as lower session IDs
> always get priority. WriteLock is not thread-safe in the first place, so
> having session ID involved in compare operation does not seem to make sense.
> (2) if findPrefixInChildren finds previous ID, it should add dir in front of
> the ID
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
