hi,
potentially troublesome due to the extra contention and possible
conflicts on the commitinfo node.
Conflicts at that point cannot occur within a single cluster node since
we are already serialised here.
this may be correct today, but I'm working on the MongoMK to support
concurrent
On 17.10.13 10:58 , Thomas Mueller wrote:
I suggest to change the mechanism. At least each cluster node should use a
different node name.
I'm on it.
Michael
Hi,
On Wed, Oct 16, 2013 at 2:58 PM, Michael Dürig mdue...@apache.org wrote:
Well there is the extra benefit for transporting that info across cluster
nodes.
I don't see the benefit. There's no reliable way to match such
external commitinfos to the respective content changes, so we can't
use
On 16.10.13 9:45 , Jukka Zitting wrote:
Yes this should work (even though I'm not too fond of using thread locals).
An cleaner alternative to a ThreadLocal would be to use an extra
commitinfo argument to carry the information down the stack through
the relevant commit and merge methods