On Fri, Mar 4, 2011 at 7:32 AM, David Jencks <[email protected]> wrote:
> I haven't found out how to start more than one feature at once.  I would 
> think that having some way to start a  feature asynchronously would be 
> useful, but I think adding a system beyond startlevel to determine bundle 
> start order is not likely to work well.
>
> When would it be a problem to start a bunch of features aysnchronously at 
> once?

Good question :) But it would definitely help in some situations to
speed up the initial startup.

Kind regards,
Andreas

>
> thanks
> david jencks
>
> On Mar 3, 2011, at 8:18 PM, Andreas Pieber wrote:
>
>> OK, another feature proposal for 3.0:
>>
>> I've the following "problem" with the OpenEngSB project right now. We
>> have different features where some are optional, but if they are
>> defined they could also start in parallel. E.g. assume you have
>> feature A and feature B. Feature A requires some time to come up since
>> it requires some time to get all blueprint services wired, BUT feature
>> B does not need to wait till feature A is finished but could rather
>> startup with feature A (as if they're defined in the same feature).
>>
>> I can think of two solutions for the problem
>>
>> 1) Define parallel startup in features.cfg. In Archlinux you have a
>> Daemons array defined saying which services to start during the boot
>> which looks something like:
>>
>> DAEMONS=(syslog-ng netfs crond dbus !network wicd @acpid @cups @ntpdate @mpd)
>>
>> This means now: first start syslo.. then netfs, then crond (one after
>> the other, similar to our feature file), but you don't have to wait
>> for acpid, cups, ntpdate to wait, start them up and go ahead (thats
>> what the @means).
>>
>> I can think of the same for Karaf:
>>
>> A,@B
>>
>> which could mean: start feature A and feature B like they where the same 
>> feature
>>
>> 2) A different solution could be to add a new tag to the features.xml
>> listing the features with which a feature could be started in parallel
>> --->
>>
>> Feature B could contain:
>>
>> <couldBeStartedTogetherWith><feature version...>A</feature><feature
>> version...>OTHER_FEATURE</feature></couldBeStartedTogetherWith>
>>
>> ------------
>> I think it may be good to support borth options since the feature
>> developer should now best with which features his feature could start
>> in parallel, BUT the feature developer can not imagine each possible
>> feature which could be started in parallel with his feature.
>>
>> So, WDYT? :)
>>
>> Kind regards,
>> Andreas
>
>

Reply via email to