Hi Mani,
The implementation of BoundedReadFromUnboundedSource currently doesn't
allow that in the same way sources checkpoint. We would either have to
convert it into a proper source (it's a DoFn atm), or store the
checkpoint mark in Beam managed state (which will be checkpointed). The
latter
Yes my issue is the lag increasing. We are using Spark Runner. Source is
Kafka and Sink is Cassandra. We tune the batch interval and max records per
batch but the batch interval still less than the processing time of each
batch. So it causes the latency. We tried to apply Reshuffle withRandomKey
on