Compatibility wise we have no other option other than deprecate. The only question I have is if we should log.warn when those configuration are in use.
To what terms primary and replica sounds good to me. On Sat, Jul 25, 2020 at 8:58 AM Christopher Shannon < christopher.l.shan...@gmail.com> wrote: > As I stated before I think this is a positive change and aligns with what > many other projects and companies are already doing. I am willing to help > out when I can and I think some others would help too. But in my opinion I > think we need a consensus on the plan before we should do anything. There > should be an agreement on what terms will be changed (and to what), how we > will handle backwards compatibility and deprecation/removal going forward, > etc. > > So far I've seen a lot of negative feedback about this proposal (more so on > the private list). I would hate to see people volunteer and spend a lot of > time making changes only to have a binding -1 vote come in which would veto > a code change. > > On Fri, Jul 24, 2020 at 3:35 PM Clebert Suconic <clebert.suco...@gmail.com > > > wrote: > > > The replacement of these becomes a heavy lift work as we need to keep > > compatibility and deprecate old terms. > > > > We need volunteers. > > > > > > <Friday joke>can we offer commit status to anyone doing it?</Friday joke> > > > > On Fri, Jul 24, 2020 at 1:51 PM Matt Pavlovich <mattr...@gmail.com> > wrote: > > > > > Chiming in on the suggestions for terms— using numeric terms (primary, > > > secondary, etc) > > > Is inconsistent since there may be multiple failover nodes that take > over > > > for the primary, > > > and it is generally non-deterministic. > > > > > > IMO having separate terms for nodes that take over a datastore and for > > > nodes that receive > > > replicated data would be a good thing because they are different > things. > > > This would allow > > > the full truth table to be indicated at any given time. > > > > > > For example: > > > master (1 node) / slave (n nodes) becomes: active (1 node) / standby > (n > > > nodes) > > > primary (1 node) -> replica (n nodes) > > > > > > > > > With this terminology, at a given time a node could be one of: > > > > > > ‘active’+‘primary’ > > > ‘active’+‘replica’ > > > ’standby’+’primary’ > > > ’standby’+’replica’ > > > > > > -Matt Pavlovich > > > > > > > On Jul 14, 2020, at 10:12 AM, Clebert Suconic < > > clebert.suco...@gmail.com> > > > wrote: > > > > > > > > I would Prefer avoiding passive and active. > > > > > > > > > > > > TBH master and slave wouldn’t offend me as a robot could be > considered > > a > > > > slave without being offensive. > > > > > > > > But if there is general consensus on the term I will leave my > personal > > > > opinion to the side there. > > > > > > > > On Tue, Jul 14, 2020 at 10:42 AM Justin Bertram <jbert...@apache.org > > > > > wrote: > > > > > > > >> Dom, internally in Artemis the process of starting the broker is > > > generally > > > >> called "activation". Therefore I typically use the terms "active" > and > > > >> "passive" to describe the "running role" as you call it. It's not > > > perfect, > > > >> but it covers most cases. > > > >> > > > >> > > > >> Justin > > > >> > > > >> On Tue, Jul 14, 2020 at 6:58 AM Domenico Francesco Bruscino < > > > >> bruscin...@gmail.com> wrote: > > > >> > > > >>> I would propose to replace `master/slave` with `leader/follower` or > > > other > > > >>> terms different from `live/backup` in ActiveMQ Artemis to keep the > HA > > > >>> configuration role of the broker separated from the HA running role > > of > > > >> the > > > >>> broker. > > > >>> For example, a broker instance with the `slave` HA configuration > role > > > >> could > > > >>> acquire the `live` HA running role after a failover. > > > >>> > > > >>> Il giorno mar 14 lug 2020 alle ore 13:42 Jiri Daněk < > > jda...@redhat.com > > > > > > > >> ha > > > >>> scritto: > > > >>> > > > >>>> On Tue, Jul 14, 2020 at 1:02 PM Xeno Amess <xenoam...@gmail.com> > > > >> wrote: > > > >>>> > > > >>>>> Like I said, I think "worker" can fully replace "slave" in every > > > >> usage > > > >>> in > > > >>>>> activeMQ. > > > >>>>> > > > >>>> > > > >>>> Nope, "worker" does not capture the idea. In Artemis, slave is > > > >>> replicating > > > >>>> the data on the master and replaces the master in case the master > > > dies. > > > >>> The > > > >>>> "worker" terminology is more suitable for a situation when the > > master > > > >>>> coordinates and all work is done on slaves. > > > >>>> > > > >>>> Looking at > > > >>>> > > > >> > > https://www.kernel.org/doc/html/latest/process/coding-style.html#naming > > > , > > > >>>> I'd suggest one of > > > >>>> > > > >>>> ‘{primary,main} / {secondary,replica,subordinate}’ ‘leader / > > follower’ > > > >>>> > > > >>>> I like the leader/follower, personally. I have a feeling I heard > it > > > >>>> somewhere in the context of database replication. > > > >>>> > > > >>>> Live / backup sounds good as well, except that "live" brings a bit > > of > > > >> the > > > >>>> echo of the notorious Unix cruelty and violence (killing children, > > > >>> reaping > > > >>>> zombies). > > > >>>> -- > > > >>>> Mit freundlichen Grüßen / Kind regards > > > >>>> Jiri Daněk > > > >>>> > > > >>> > > > >> > > > > -- > > > > Clebert Suconic > > > > > > -- > > Clebert Suconic > > > -- Clebert Suconic
