It seems the problem is just between kafka 0.10.x --> kafka 2.x. It is working fine for kafka 2.x --> 2.x.
On Sat, Oct 3, 2020 at 4:50 PM Samuel Cantero <scante...@gmail.com> wrote: > Hi all, > > It looks like the offsets returned by the > RemoteClusterUtils.translateOffsets are always the same as the > source/primary cluster, i.e, there's no actual translation. > > The upstream offsets are fine but the downstream are not (it is just the > same as the upstream one). After a quick look, it seems the bug might be on > the writer of the source-cluster-alias.checkpoints.internal topic rather > than on this lib that decodes it. > > I have double checked the log-end-offset of the topic on the target > cluster via > ./kafka-run-class.sh kafka.tools.GetOffsetShell --broker-list > KAFKA-CONN-STR --time -1 --topic TOPIC > > This gives me confirmation that there's no actual translation. > > To double check it is not just my app, I enabled > sync.group.offsets.enabled > <https://cwiki.apache.org/confluence/display/KAFKA/KIP-545%3A+support+automated+consumer+offset+sync+across+clusters+in+MM+2.0> > and > I can find the same problem. So I think there's a bug unless I'm doing > something wrong. > > I'm using the latest changes from trunk. Anyone hitting the same issue? > > Best, >
