I thought the fix on the master is not changing the default behavior. If it is, then regardless of which approach we take we will have a different behavior. Hence we will need to release note in either case.
On Wed, May 11, 2016 at 9:46 AM, Luis Gomez <[email protected]> wrote: > Default behavior an small API change as far as I remember, that is why we > decide to do it in master only. > > On May 11, 2016, at 12:50 AM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Yes we have fix on master, but I think we all agreed put this fix only on > master because it change the default behavior. > > Jozef > > *From:* Abhijit Kumbhare [mailto:[email protected] > <[email protected]>] > *Sent:* 11. mája 2016 8:33 > *To:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco) < > [email protected]> > *Cc:* Anil Vishnoi <[email protected]>; [email protected]; > Abhijit Kumbhare <[email protected]>; Subhash Singh < > [email protected]>; Luis Gomez Palacios < > [email protected]>; [email protected] > *Subject:* Re: [openflowplugin-dev] table features > > The next SR is July end - and I believe you will have the actual fix in > the master & not a workaround - right? If that is the case - then can that > be ported to stable/beryllium before the next SR? > > On Tue, May 10, 2016 at 11:01 PM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Abhijit hard to decide but if we declare some performance in SR2 we should *if > possible* put it in this release. And the pull from master you mean the > actual fix yang model structure change not this "workaround" ? > > > > Jozef > ------------------------------ > *From:* Abhijit Kumbhare <[email protected]> > *Sent:* Tuesday, May 10, 2016 5:32 PM > *To:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco) > *Cc:* Anil Vishnoi; [email protected]; Abhijit Kumbhare; Subhash > Singh; Luis Gomez Palacios; [email protected] > > *Subject:* Re: [openflowplugin-dev] table features > > Jozef, > > On a second glance - the patch is not merged. So we will not need to add > it to the SR2 release notes (the build done on Friday). So can you please > update your patch to have the following description so we will not forget > when it is time to actually release note? > > Added the ability to configure whether to pull table features for the Li > design and changed the default to skip pulling table features for the > feature. > Data yang model defines a table features inside of table grouping. OVS 2.4 > finally supports table features. Now large table features data are stored > inside of each table. It means 254 table features are stored in DS. This > ability allows skip pulling and storing of large table features. Table > features are still available via rpc but if set to true then maintenance in > DS will be omitted and DS latency for inventory will be the same as by OVS > >=2.3. > > Secondly - if the table features performance fix is already available & > since the next SR is a few months away - do you think it would be better to > just pull the actual fix from master once we know there are no side effects? > > Thanks, > Abhijit > > On Tue, May 10, 2016 at 7:22 AM, Abhijit Kumbhare <[email protected]> > wrote: > > Thanks Jozef! Good explanation - I will work with An later today to get it > into the release notes. > > On Tue, May 10, 2016 at 4:05 AM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Data yang model defines a table features inside of table grouping. OVS 2.4 > finally supports table features. Now large table features data are stored > inside of each table. It means 254 table features are stored in DS. This > ability allow skip pulling and storing of large table features. Table > features are still available via rpc but if set to true then maintenance in > DS will be omitted and DS latency for inventory will be the same as by OVS > >=2.3. > > > ------------------------------ > *From:* Anil Vishnoi <[email protected]> > *Sent:* Tuesday, May 10, 2016 9:44 AM > > *To:* Abhijit Kumbhare > *Cc:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco); > [email protected]; Abhijit Kumbhare; Subhash Singh; > [email protected]; Luis Gomez Palacios > *Subject:* Re: [openflowplugin-dev] table features > > > > On Tue, May 10, 2016 at 12:30 AM, Abhijit Kumbhare <[email protected]> > wrote: > > This change was due to the major performance hit on OVS 2.4 - and I > believe was a decision at one of the meetings. Reason: the inefficient > table features fetch were always causing a significant performance drop > regardless of whether DIDM or any other features needing table features > were enabled or not. If I remember right - the decision was to turn off the > table features on stable/beryllium & add the riskier fix on the master > first & then maybe port it to the stable/beryllium after some time it has > been baked in (may be in the next SR). > > So it would be better to release note it I think. > > Yes, so lets do it. > > > > Do you have a short 3-4 line description Jozef for release note > (explaining why it was changed)? > > On Tue, May 10, 2016 at 12:22 AM, Anil Vishnoi <[email protected]> > wrote: > > Personally i want it to be disabled by default, but if i look at it from > user perspective, we are changing the behaviour between two SR version and > it might be of some concern. But if this change is for lithium plugin, i > think the impact is minimal, so i think if we can release note it, that > would be better. > > On Mon, May 9, 2016 at 11:32 PM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > You propose to leave the flag (skip table features) false ? Just not to > change default behavior? > > > > Jozef > ------------------------------ > *From:* Anil Vishnoi <[email protected]> > *Sent:* Monday, May 9, 2016 7:21 PM > *To:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco) > *Cc:* Abhijit Kumbhare; [email protected]; Abhijit Kumbhare; Subhash > Singh; [email protected]; Luis Gomez Palacios > > *Subject:* Re: [openflowplugin-dev] table features > > okay, although it's been done for li plugin, but i think it will change > the default behavior between two SR (SR2 and SR-3). > > On Mon, May 9, 2016 at 1:48 AM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Nope, in stable/beryllium we add just this on/off flag, with default > setting on OFF table features. > > > > https://git.opendaylight.org/gerrit/#/c/36506/3 > > > > Jozef > ------------------------------ > *From:* Anil Vishnoi <[email protected]> > *Sent:* Monday, May 9, 2016 10:37 AM > *To:* Abhijit Kumbhare > *Cc:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco); > [email protected]; Abhijit Kumbhare; Subhash Singh; > [email protected]; Luis Gomez Palacios > > *Subject:* Re: [openflowplugin-dev] table features > > was this patch merged to stable/beryllium as well? > > On Tue, Apr 26, 2016 at 8:21 AM, Abhijit Kumbhare <[email protected]> > wrote: > > OK. > > On Tue, Apr 26, 2016 at 12:18 AM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Hi Abhijit, > > > > I thought I get an answer from NIC and DIDM guys, but in this case, I > would propose we just make the on/off flag in beryllium SR3 and this > solution we merge only into master M3 as we agreed with Luiz. > > > > Jozef > ------------------------------ > *From:* Abhijit Kumbhare <[email protected]> > *Sent:* Monday, April 25, 2016 9:20 PM > *To:* Subhash Singh > *Cc:* Jozef Bacigal -X (jbacigal - PANTHEON TECHNOLOGIES at Cisco); > Abhijit Kumbhare;[email protected]; > [email protected] > *Subject:* Re: [openflowplugin-dev] table features > > Hi Jozef, > > Will you be putting https://git.opendaylight.org/gerrit/#/c/36559 into > stable/beryllium after sometime has passed or do you think its better to > avoid it altogether in stable/beryllium? > > Thanks, > Abhijit > > On Mon, Apr 25, 2016 at 5:33 AM, Subhash Singh < > [email protected]> wrote: > > +[Anandhi] > > -- > Regards, > Subhash Kumar Singh > > On Mon, Apr 25, 2016 at 2:42 PM, Jozef Bacigal -X (jbacigal - PANTHEON > TECHNOLOGIES at Cisco) <[email protected]> wrote: > > Hi everyone, mainly guys from NIC and DIDM, > > I would ask you if you can read and talk about the bug 5464 table features > > https://bugs.opendaylight.org/show_bug.cgi?id=5464 > > There are two proposals, first (berylium SR2) that we merge the skip flag, > which is I would say some "workaround" and set the flag to TRUE so we > default skip the table features > > https://git.opendaylight.org/gerrit/#/c/36506/ > > and second (boron M3) this solution of the problem which would lead to > changes into your projects > > https://git.opendaylight.org/gerrit/#/c/36559 > > We would much appreciate you answers > > Jozef > > > > > _______________________________________________ > openflowplugin-dev mailing list > [email protected] > https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev > > > > > > > _______________________________________________ > openflowplugin-dev mailing list > [email protected] > https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev > > > > > -- > Thanks > Anil > > > > > -- > Thanks > Anil > > > > > -- > Thanks > Anil > > > > > > > -- > Thanks > Anil > > >
_______________________________________________ openflowplugin-dev mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
