No, we do not guarantee that an entry will not be lost. When the split happens, either one partition could end up with 0 owners for a particular segment, and it will allocate new owners for that segment. When the merge happens, that partition's consistent hash may be chosen as the merge consistent hash, and keys on the other partition's owners (which could still have the value) will be ignored.

Cheers

Dan

On Wed, Jan 21, 2015 at 10:48 AM, Radim Vansa <rvansa@redhat.com> wrote:

Hi,

one question on the forum [1] led me to thinking whether we offer any
guarantees at all after merge *without* partition handling. The common
sense would suggest that we could have inconsistency on entries
overwritten in one of the partitions, but the entry should not be lost
completely.

Do we have at least some unit tests trying to confirm this? I have to
admit that I was not testing this scenario with RadarGun.

Radim

[1]
https://developer.jboss.org/message/916484?et=watches.email.thread#916484

--
Radim Vansa <rvansa@redhat.com>
JBoss DataGrid QA

_______________________________________________
infinispan-dev mailing list
infinispan-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev