On Mon, Dec 11, 2017 at 5:28 PM, Michael Vorburger <[email protected]>
wrote:

> On Mon, Dec 11, 2017 at 8:36 PM, Tom Pantelis <[email protected]>
> wrote:
>
>> On Mon, Dec 11, 2017 at 2:22 PM, Michael Vorburger <[email protected]>
>> wrote:
>>
>>> Controllers,
>>>
>>> looking at https://jira.opendaylight.org/browse/NETVIRT-916, half of
>>> the solution could be https://git.opendaylight.org/gerrit/#/c/66355/,
>>> and am wondering if the other half of the solution is in controller's
>>> ConcurrentDOMDataBroker:
>>>
>>> Is it right for it to LOG.warn a ConflictingModificationAppliedException?
>>> Shouldn't that be left to the caller? Given that a failed Future is
>>> returned, why log it from controller? Because people could just ignore the
>>> Future? IMHO we now have a solution for this via the @CheckReturnValue
>>> which error-prone (and maybe FindBugs, I'm not sure) can verify. Only for
>>> projects enforcing such tools, of course. And only if we
>>> add @CheckReturnValue to WriteTransaction submit() which I think we should
>>> - OK for everyone?
>>>
>>> Similarly for an OptimisticLockFailedException (not the case in
>>> NETVIRT-916, but just while we're at it) - that IMHO also should be be WARN
>>> logged by controller (if it currently is; dunno).
>>>
>>
>> Yeah I think that was put in case callers ignore the returned Future. I
>> agree it mostly adds a lot of extra noise - I'm fine with lowering it to
>> DEBUG.
>>
>
> OK, great; then I've just so proposed this in a few changes on
> https://git.opendaylight.org/gerrit/#/q/topic:CONTROLLER-1802
>
> What's may be still TBD is the equivalent of https://git.opendaylight.org/
> gerrit/#/c/66362/ in mdsal, or is there no such thing?
>

ConcurrentDOMDataBroker is CDS's broker implementation
(sal-distributed-datastore) and is the one used in production.
_______________________________________________
controller-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/controller-dev

Reply via email to