Hi Alex/Praveen, I don't remember being part of a discussion where the STANDBY assignments should be given the same treatment as 'QUIESCED/QUIESCING' in this scenario. I guess the implementation just took that route based on the words "or another HA state" as-in P 186, section 3.8.1.3, lines29- 30 - "the active HA state is removed from components or another HA state is assigned to components"
In my opinion, I think both ACTIVE and STANDBY assignments should follow the same ordering of dependencies . I also think this brings no additional considerations whether it is PI or NPI. Cheers, Mathi. >-----Original Message----- >From: Alex Jones [mailto:[email protected]] >Sent: Thursday, March 31, 2016 7:24 PM >To: praveen malviya; [email protected] >Subject: Re: [users] AMF question on CSI ordering > >Thanks Praveen. > >Maybe I need to be a little clearer here. > >We have two SA-aware components in the same SI. We use >saAmfCSIDependencies to make sure that the ACTIVE assignment for one CSI >always precedes the other. This works as we expect. But, we notice that it is >not the case when the STANDBY assignment is done. And I don't understand >why. I understand for QUIESCED/QUIESCING that the order should be >reversed, but why not have the same order for STANDBY as ACTIVE for SA- >aware components? > >I don't see a mention of NPI anywhere in 3.8.1.3, so I'm not sure what you >mean here. > >Alex > > >On 03/30/2016 09:18 AM, praveen malviya wrote: >> Please see inline. >> >> Thanks, >> Praveen >> >> On 30-Mar-16 6:18 PM, Alex Jones wrote: >>> Hi, >>> >>> Does anyone know what is the reasoning behind "*/reverse order >>> is applied when /**/... another HA state is assigned to components/*" >>> in the AMF spec in 3.8.1.3. Apparently OpenSAF interprets this for >>> STANDBY assignment. In other words, if I use CSIDependencies I can >>> order the ACTIVE assignment for each CSI, but the STANDBY CSI >>> assignments are done in reverse. I understand why QUIESCING or >>> QUIESCED should be done in reverse order, but why STANDBY? I would >>> think you would want the ordering the same as for ACTIVE. >>> >>> Can someone explain this? >> >> In the section 3.8.1.3, the example used for explanation of CSI deps >> models NPI components. In the next section (3.8.2 page 188), spec >> clearly differentiate between the use of saAmfCompinstantiationlevel >> and CSI deps for resolving dependencies among components. >> So in case of NPI components, there will be no assignments for >> standby. So the question of order becomes invalid. For >> quiesced/quiescing state termination will be done in reverse order and >> it goes with spec. >> >> At the same time for sa-aware compnents, SI deps is the way that spec >> proposes for any assignment related dependencies and >> saAmfCompinstantiationlevel for any execution environment related deps. >>> >>> Alex >>> >>> --------------------------------------------------------------------- >>> --------- >>> >>> Transform Data into Opportunity. >>> Accelerate data analysis in your applications with Intel Data >>> Analytics Acceleration Library. >>> Click to learn more. >>> http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140 >>> _______________________________________________ >>> Opensaf-users mailing list >>> [email protected] >>> https://lists.sourceforge.net/lists/listinfo/opensaf-users >>> >> > > >------------------------------------------------------------------------------ >Transform Data into Opportunity. >Accelerate data analysis in your applications with Intel Data Analytics >Acceleration Library. >Click to learn more. >http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140 >_______________________________________________ >Opensaf-users mailing list >[email protected] >https://lists.sourceforge.net/lists/listinfo/opensaf-users ------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140 _______________________________________________ Opensaf-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/opensaf-users
