sara at sinodun.com
Fri Aug 24 14:19:12 UTC 2012
On 24 Aug 2012, at 14:23, Jerry Lundström wrote:
> On Aug 24, 2012, at 15:12 , Sara Dickinson wrote:
>>> The reason is singular as far as I can tell. Zones could share keys.
>>> Thus creating new keys / deciding to reuse an existing key creates a
>>> race condition for those policies.
>> So how does the multi-threaded enforcer avoid this problem?
> With locks when doing stuff with shared keys policies and database transactions.
You agree with Yuri then :-) It would be nicer to have the same solution in both releases but if there may be reasons this can't be done.
Sounds like something to discuss at the developer workshop?
>>>> 2. Enabling of multithreaded Enforcer
>> +1; although leave the compile flag in, in case folk want to build without it?
>> It also means that package maintainers will not need to build two versions (or decide on one or the other)
>> Are we going to make the "enforcer workers" tag mandatory? I vote for optional and if it doesn't exist then workers == 1.
> We can easily redo it to --disable-enforcer-workers.
A --disable-enforcer-workers compile flag sound like the right answer to me. Agree with Sion on tag handling.
More information about the Opendnssec-develop