[infinispan-issues] [JBoss JIRA] (ISPN-3357) Insufficient owners with putIfAbsent during node join rebalance

[Dan Berindei (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-3357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12793625#comment-12793625 ] 

Dan Berindei commented on ISPN-3357:
------------------------------------

[~nekop] You didn't mention the transaction mode in the description, but from the logs this looks like a non-tx cache.

[~mircea.markus] Right, a backup owners should not check the previous value for conditional operations, only the primary owner should do it. The primary should also set the ignorePreviousValue flag on the command before forwarding it to the backup owners.

State transfer-related forwarding could be done only from the primary owner as well for non-tx caches. The forwarding would still be done from the StateTransferInterceptor, so without holding the primary lock, and without guaranteeing consistency, but I don't think we could lose updates either.

For tx caches, we already set the {{ignorePreviousValue}} flag before replicating the command to the owners (via a PrepareCommand). But we reset it when we process the command on the owners, and I don't think that's ok because the PrepareCommand could be forwarded by StateTransferInterceptor from one of the owners, with {{ignorePreviousValue == false}}.
                
> Insufficient owners with putIfAbsent during node join rebalance
> ---------------------------------------------------------------
>
>                 Key: ISPN-3357
>                 URL: https://issues.jboss.org/browse/ISPN-3357
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Distributed Cache
>    Affects Versions: 5.2.4.Final, 6.0.0.Alpha1
>            Reporter: Takayoshi Kimura
>            Assignee: Dan Berindei
>            Priority: Critical
>         Attachments: 7c29bccb.log
>
>
> Here is test scenario:
> * DIST numOwners=2, start with 3 nodes cluster then join 1 node during load
> * HotRod putIfAbsent accesses from 40 threads (1 process, 1 remote cache instance), 40000 entries total
> After the test run, the numberOfEntries on each node are:
> * node1: 20074
> * node2: 19888
> * node3: 20114
> * node4: 18885
> Total is 78961, 1039 entries are missing. No error on HotRod client side so 80000 entries should be there.
> Let's take a look at example missing entry, hash(thread01key151) = 7c29bccb.
> Current CH: owners(7c29bccb) are [node1, node2]
> Pending CH: owners(7c29bccb) are [node1, node2, node4]
> Balanced CH: owners(7c29bccb) are [node1, node4]
> The events sequence is:
> * hotrod -> node1
> * node1 -> node2, node4
> * node2 committed entry
> * node4 performed clustered get before write, got a value from node2 and will not commit the entry because this node thinks it's not changed/created
> * node1 committed entry
> * node2 invalidates the entry because it's no longer an owner
> Result owners(7c29bccb) are only node1 and node4 is missing. This entry may be completely lost by further rebalances when node4 is donor for this segment.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira