On 03/12/10 08:01 AM, Sebastien Roy wrote:
> On 03/12/10 09:41 AM, Anurag S. Maskey wrote:
>> nwam phase 1 included hacks to work around the fact that when nwamd
>> starts up on the first reboot, it is running on the old manifest. These
>> hacks have to be removed since with Early Manifest Import these are not
>> required. Either the hack have to removed with this project or bugs
>> filed to remove this hack. I wonder if there are other projects/code
>> that have introduced similar workarounds. Ignore this message if this
>> has already been worked out.
>
> It would likely help the project team if the specific hacks were 
> pointed out so that they could fix them.  One that I'm personally 
> aware of is in /lib/svc/share/net_include.sh's net_reconfigure function:
>
>         #
>         # Ensure that the datalink-management service is running since
>         # manifest-import has not yet run for a first boot after
>         # upgrade.  We wouldn't need to do that if manifest-import ran
>         # earlier in boot, since there is an explicit dependency
>         # between datalink-management and network/physical.
>         #
>         svcadm enable -ts network/datalink-management:default
>
> There's an entire block of related code below this svcadm incantation 
> that also needs to be removed.
>
> I ran a quick find in /lib/svc grepping for manifest-import, and found 
> that the ipfilter method script has an entire diatribe in its comments 
> section dedicated to the manifest-import problem.  Someone needs to go 
> look into what that's all about and change it appropriately.
>
> -Seb
> _______________________________________________
> smf-discuss mailing list
> smf-discuss at opensolaris.org
>
>

At this time we will check the methods delivered for any obvious "hacks" 
and file bugs for the providers of the service to remove/fix the hacks 
to take advantage of Early Manifest Import, if bugs have not already 
been filed.  It does not appear that we are breaking any of the services 
that use these hacks (especially if they check for the expected state 
before applying the new state, the state will already be there and there 
should be no new state application).



-- 
Sean Wilcox
303.272.9711
x79711

Reply via email to