Thanks again to everyone for all the work on this KIP and implementation! I've discovered that it would be easier for downstream projects if the new `SinkTaskContext.errantRecordReporter()` method were a default method that returns null. Strictly speaking it's not required as Connect will provide the implementation for the Connect runtime, but some downstream projects may use their own implementations of this interface for testing purposes. See https://issues.apache.org/jira/browse/KAFKA-10111 for details and https://github.com/apache/kafka/pull/8814 for the suggested change. IMO there is little harm in making the existing non-default method a default that returns null, but please let me know if you object.
Best regards, Randall On Thu, May 21, 2020 at 2:10 PM Randall Hauch <rha...@gmail.com> wrote: > The vote has been open for >72 hours, and the KIP is adopted with three +1 > binding votes (Konstantine, Ewen, me), one +1 non-binding vote (Andrew), > and no -1 votes. > > I'll update the KIP and the AK 2.6.0 plan. > > Thanks, everyone. > > On Tue, May 19, 2020 at 4:33 PM Konstantine Karantasis < > konstant...@confluent.io> wrote: > >> +1 (binding) >> >> I like how the KIP looks now too. Quite active discussions within the past >> few days, which I found very useful. >> >> There's some room to allow in the future the connector developers to >> decide >> whether they want greater control over error reporting or they want the >> framework to keep providing the reasonable guarantees that this KIP now >> describes. The API is expressive enough to accommodate such improvements >> if >> they are warranted, but its current form seems quite adequate to support >> efficient end-to-end error reporting for sink connectors. >> >> Thanks for introducing this KIP Aakash! >> >> One last minor comment around naming: >> Currently both the names ErrantRecordReporter and failedRecordReporter are >> used. Using the same name everywhere seems preferable, so feel free to >> choose the one that you prefer. >> >> Regards, >> Konstantine >> >> On Tue, May 19, 2020 at 2:30 PM Ewen Cheslack-Postava <e...@confluent.io> >> wrote: >> >> > +1 (binding) >> > >> > This will be a nice improvement. From the discussion thread it's clear >> this >> > is tricky to get right, nice work! >> > >> > On Tue, May 19, 2020 at 8:16 AM Andrew Schofield < >> > andrew_schofi...@live.com> >> > wrote: >> > >> > > +1 (non-binding) >> > > >> > > This is now looking very nice. >> > > >> > > Andrew Schofield >> > > >> > > On 19/05/2020, 16:11, "Randall Hauch" <rha...@gmail.com> wrote: >> > > >> > > Thank you, Aakash, for putting together this KIP and shepherding >> the >> > > discussion. Also, many thanks to all those that participated in >> the >> > > very >> > > active discussion. I'm actually very happy with the current >> proposal, >> > > am >> > > confident that it is a valuable improvement to the Connect >> framework, >> > > and >> > > know that it will be instrumental in making sink tasks easily >> able to >> > > report problematic records and keep running. >> > > >> > > +1 (binding) >> > > >> > > Best regards, >> > > >> > > Randall >> > > >> > > On Sun, May 17, 2020 at 6:59 PM Aakash Shah <as...@confluent.io> >> > > wrote: >> > > >> > > > Hello all, >> > > > >> > > > I'd like to open a vote for KIP-610: >> > > > >> > > > >> > > >> > >> https://cwiki.apache.org/confluence/display/KAFKA/KIP-610%3A+Error+Reporting+in+Sink+Connectors >> > > > >> > > > Thanks, >> > > > Aakash >> > > > >> > > >> > > >> > >> >
