Hey, sorry for dropping the ball on this. Your reasoning makes sense, I'll go over to the vote thread now, :)
On Tue, Sep 24, 2024 at 9:03 AM Dániel Urbán <urb.dani...@gmail.com> wrote: > Bump - I'll start a vote soon if there are no further comments. > Daniel > > Dániel Urbán <urb.dani...@gmail.com> ezt írta (időpont: 2024. szept. 19., > Cs, 12:23): > > > Bump - I know this is not a complicated proposal, but if you have any > > inputs, please share. > > TIA, > > Daniel > > > > Dániel Urbán <urb.dani...@gmail.com> ezt írta (időpont: 2024. szept. > 13., > > P, 16:54): > > > >> Hi Viktor, > >> Thanks for your comments. Technically, yes, a policy subclass can be > used > >> to intentionally misconfigure the heartbeats topic name. Mind you, this > >> still does not disable the replication of the heartbeats topics, simply > the > >> Connector will not be able to those topics. > >> > >> VS1. Only in connector mode. Dedicated mode does not allow multiple > >> MirrorSourceConnector instances in a single flow, so the issue is not > >> present. > >> VS2. I assume you mean MirrorSourceConnectors instead of heartbeat > >> connectors. As mentioned in the alternatives section, the KIP-1016 > >> configuration could be used, but again, it still does not disable the > >> replication, just "misconfigures" the connector to not be able to find > the > >> heartbeats topics. It can be a workaround to the issue described in the > >> motivation, but I consider it to be more like a hack than a proper > >> workaround. > >> > >> Thanks, > >> Daniel > >> > >> Viktor Somogyi-Vass <viktor.somo...@cloudera.com.invalid> ezt írta > >> (időpont: 2024. szept. 13., P, 16:41): > >> > >>> Hi Daniel, > >>> > >>> As you describe the case, I think it could be a good idea potentially > to > >>> add this, however we have to note that customers can achieve this > >>> behavior > >>> today by subclassing the default policy. This is more cumbersome though > >>> than flipping a boolean though, perhaps we could add this to the > rejected > >>> alternatives. > >>> > >>> Some questions: > >>> VS1. Does this problem exist in both deployment cases (dedicated vs > >>> connector modes)? > >>> VS2. Regarding KIP-1016: wouldn't it help if given that KIP, one > deployed > >>> multiple heartbeat connectors with different names and compatible > >>> filters? > >>> If one could specify the heartbeat topic name for only the given > >>> connector, > >>> then it would ignore the others (but maybe there is a mistake in my > >>> thinking). > >>> > >>> Thanks, > >>> Viktor > >>> > >>> On Thu, Sep 12, 2024 at 4:26 PM Dániel Urbán <urb.dani...@gmail.com> > >>> wrote: > >>> > >>> > Hi everyone, > >>> > I'd like to kick off the discussion of a small improvement in the > >>> > MirrorSourceConnector: > >>> > > >>> > > >>> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-1089%3A+Allow+disabling+heartbeats+replication+in+MirrorSourceConnector > >>> > The change would be fairly simple, and enables some more advanced > >>> use-cases > >>> > of the Connector. > >>> > TIA, > >>> > Daniel > >>> > > >>> > >> >
