[
https://issues.jboss.org/browse/ISPN-2472?page=com.atlassian.jira.plugin....
]
Dan Berindei updated ISPN-2472:
-------------------------------
Status: Pull Request Sent (was: Open)
Git Pull Request:
https://github.com/infinispan/infinispan/pull/1469
It turns out it's enough if the originator also checks the command's topology and
forwards it to the new owners.
Because the prepare needs to be successful on all nodes, it doesn't matter if the
originator really acquired the lock - it's enough that the original primary owner
did.
There is a catch though - because of ISPN-2510, if the cache on the primary owner is
stopped, but the cache manager is still running, we report success. We should instead
report failure and either fail or retry on the originator (see ISPN-2402).
Transactional commands should not be forwarded to the tx originator
-------------------------------------------------------------------
Key: ISPN-2472
URL:
https://issues.jboss.org/browse/ISPN-2472
Project: Infinispan
Issue Type: Sub-task
Components: State transfer, Transactions
Affects Versions: 5.2.0.Beta3
Reporter: Dan Berindei
Assignee: Dan Berindei
Priority: Critical
Fix For: 5.2.0.Beta4
As discussed in the parent issue (ISPN-2410), transactional commands should not be
forwarded back to the tx originator.
Instead, when the RPC returns, the originator should check if it needs to lock or commit
extra keys (because there was a rebalance during the RPC) and do that work itself.
--
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