FWIW here's the thread from april 15 where we talked about this at length, for some reference points:
http://mail-archives.apache.org/mod_mbox//zookeeper-dev/201504.mbox/%3CCAKF1A=ulfxttvw+gtt3gynzwtpjz60wot5zhmk2fkum2vat...@mail.gmail.com%3E On Mon, Aug 29, 2016 at 2:42 PM, Flavio Junqueira <f...@apache.org> wrote: > I'm actually not pushing back at all, I just want to make sure we do it > right rather than rush into doing it. In particular, I'd like to understand > whether the story between TTL nodes and sessions is crispy. I believe etcd > has TTL because they don't have sessions and can't have ephemerals... or at > least didn't have sessions, is it still the case? In our case, we have > chosen early on to have sessions. Having TTL nodes seems to give the option > of not relying on sessions, but if I remember correctly, this is not what > we are doing in the current patch. The client still creates a session and > issues requests through a session. I'm mostly trying to see from the > perspective of a user what I'd need to do to benefit from the feature, when > it makes sense to use it rather than ephemerals, and how to do it in a > meaningful way. > > On the server side, we already have a mechanism to expire sessions, do we > a separate scheme to expire TTL nodes or can we use the same mechanism? > Does it make sense to consider a TTL node as a degenerate case of a session > in which I have a single ephemeral node? My recollection is that it > currently uses the container manager instead. > > -Flavio > > > On 29 Aug 2016, at 19:19, Camille Fournier <cami...@apache.org> wrote: > > > > I'm happy to take back up the conversation of why have this at all. > > I proposed this feature after doing some personal research into etcd for > a > > talk about a year and a half ago. I found that one of the interesting > > features of etcd was that you could have clients enable ephemeral data > > without sticky sessions to the server. This is a popular feature in etcd > > and I can understand the desire to have a more lightweight way to create > > such nodes. > > > > I suppose I will kick it back to you, what are you afraid of vis a vis > > usage? > > > > On Mon, Aug 29, 2016 at 10:56 AM, Flavio Junqueira <f...@apache.org> > wrote: > > > >> I've left some time back a request to have it somewhere so that I could > >> leave some specific review comments and asked a question about the API > >> changes. > >> > >> I'd like to understand at a high level what we are trying to achieve. > For > >> example, the description of the jira mentions that the goal is to enable > >> nodes to expire without relying on sessions. Does it imply that this is > for > >> applications that will rely purely on local sessions? Should we provide > a > >> way of not having sessions at all, global or local? > >> > >> My sense is that this is a great feature and I'm happy to see a patch > and > >> discussion, but I feel that we need to discuss it further so that we > >> understand how this is going to be used. At least, I'd like to > understand > >> it better. > >> > >> -Flavio > >> > >>> On 29 Aug 2016, at 13:11, Jordan Zimmerman <jor...@jordanzimmerman.com > > > >> wrote: > >>> > >>> Per Benjamin: "i'm fine letting it go in as is" > >>> > >>>> On Aug 26, 2016, at 8:18 PM, Patrick Hunt <ph...@apache.org> wrote: > >>>> > >>>> Thanks Camille! It's fine with me, although notice that Flavio has > been > >>>> providing feedback and has some concerns. Also there is a pending > issue > >>>> (testing) identified by Ben most recently and afaict not yet resolved. > >>>> > >>>> Patrick > >>>> > >>>> On Fri, Aug 26, 2016 at 1:49 PM, Camille Fournier <cami...@apache.org > > > >>>> wrote: > >>>> > >>>>> OK I'm the slacker that proposed us doing this in the first place and > >> I'm > >>>>> THRILLED that you have done it Jordan, thank you so much. > >>>>> Pat, I can review and merge, unless you are concerned with > >> interference on > >>>>> other issues. LMK. > >>>>> > >>>>> C > >>>>> > >>>>> On Fri, Aug 26, 2016 at 10:42 AM, Patrick Hunt <ph...@apache.org> > >> wrote: > >>>>> > >>>>>> Hi Jordan. This looks like a great new feature, but I'm afraid I'm > >>>>> focused > >>>>>> on other things atm. I don't have much time after work/home > >> currently, as > >>>>>> such I've been focused on other priorities; 1) supporting existing > >>>>>> users/issues in 3.4, and 2) trying to get 3.5 branch to production > >> ready. > >>>>>> There are already a number of features queued up in that branch > (3.5) > >>>>> which > >>>>>> we need to get out to folks. Thanks for your patience. > >>>>>> > >>>>>> Regards, > >>>>>> > >>>>>> Patrick > >>>>>> > >>>>>> On Thu, Aug 25, 2016 at 9:00 PM, Jordan Zimmerman < > >>>>>> jor...@jordanzimmerman.com> wrote: > >>>>>> > >>>>>>> So - what’s a guy got to do to get this merged? > >>>>>>> > >>>>>>> -Jordan > >>>>>>> > >>>>>>>> On Aug 18, 2016, at 11:27 AM, Andrew Purtell <apurt...@apache.org > > > >>>>>>> wrote: > >>>>>>>> > >>>>>>>> Please > >>>>>>>> This looks really handy for implementing transient data > structures. > >>>>>>>> > >>>>>>>> On Thu, Aug 18, 2016 at 5:53 AM, Jordan Zimmerman < > >>>>>>>> jor...@jordanzimmerman.com> wrote: > >>>>>>>> > >>>>>>>>> Any chance of getting https://issues.apache.org/ > >>>>>>> jira/browse/ZOOKEEPER-2169 > >>>>>>>>> merged? It has: > >>>>>>>>> > >>>>>>>>> * A patch that’s been reviewed > >>>>>>>>> * 7 Votes > >>>>>>>>> * 15 Watchers > >>>>>>>>> * Will help ZooKeeper compete against etcd/consul > >>>>>>>>> > >>>>>>>>> -Jordan > >>>>>>>> > >>>>>>>> > >>>>>>>> > >>>>>>>> > >>>>>>>> -- > >>>>>>>> Best regards, > >>>>>>>> > >>>>>>>> - Andy > >>>>>>>> > >>>>>>>> Problems worthy of attack prove their worth by hitting back. - > Piet > >>>>>> Hein > >>>>>>>> (via Tom White) > >>>>>>> > >>>>>>> > >>>>>> > >>>>> > >>> > >> > >> > >
