[jboss-jira] [JBoss JIRA] (JGRP-2220) Two nodes can own a same lock at same time

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-2220?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13468525#comment-13468525 ] 

Bela Ban commented on JGRP-2220:
--------------------------------

I don't support such an old (4 years) version. Please try this with 3.6.x or 4.x!

I ran LockServiceDemo on master (4.0.7-SNAPSHOT):
* Start node A
* Start node B
* A: {{lock X}}: lock X is acquired successfully
* B: {{lock X}}: B blocks.
* A: {{unlock X}}: B acquires the lock
* CTRL-C A
* Restart node A
* A: {{lock X}}: this blocks until B calls unlock or is CTRL-C'ed

Note that trylock() with a long timeout is the same as lock(). Both lock methods passed in this test

> Two nodes can own a same lock at same time
> ------------------------------------------
>
>                 Key: JGRP-2220
>                 URL: https://issues.jboss.org/browse/JGRP-2220
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.4
>         Environment: Two nodes cluster with jgroup 3.4.0 alpha2
>            Reporter: Xiao Li
>            Assignee: Bela Ban
>             Fix For: 4.0.7
>
>
> We have a two node cluster with following protocol stack for distributed lock. 
> lock.protocolStack=UDP(bind_addr=myIP;bind_port=31562;mcast_addr=239.255.166.17;mcast_port=31569;ip_ttl=32;mcast_send_buf_size=150000;mcast_recv_buf_size=80000):PING(timeout=2000;num_initial_members=3):MERGE2(min_interval=5000;max_interval=10000):FD_ALL(interval=5000;timeout=20000):VERIFY_SUSPECT(timeout=1500):pbcast.NAKACK(retransmit_timeout=300,600,1200,2400,4800):UNICAST(timeout=5000):pbcast.STABLE(desired_avg_gossip=20000):FRAG(frag_size=8096):pbcast.GMS(join_timeout=5000;print_local_addr=true):CENTRAL_LOCK(num_backups=2)
> We use the  following call to get a lock for one node.
> lock = lockManager.getLock(lockId);
> lock.tryLock(Long.MAX_VALUE, TimeUnit.MILLISECONDS); where lockManager is an object of LockService and lock is a Lock object.
> Another node is waiting on the lock at same method call.  When the node owning the lock release the lock with the call lock.unlock(), the node waiting on the lock gets the lock and the lock is deleted from the node owned the lock. When the JVM for the node that released the lock restarted, it can also get same lock. So two nodes own the lock at same time. 
>   If I don't call lock.unlock() before shutting down the node, it waits for the lock as expected when its JVM starts up. 



--
This message was sent by Atlassian JIRA
(v7.2.3#72005)