I see. For normal maintenance operations, before you kill your container you could shuts down the Streams application by calling `KafkaStreams#close()`. Upon shutting down it would write a local checkpoint file indicating at which point in terms of offsets it has stopped at. So on resuming if the tasks gets allocated to the same host it can restart from that checkpoint offset position than wiping out the whole state and restarting from scratch.
Guozhang On Tue, Aug 15, 2017 at 11:36 PM, Anish Mashankar <an...@systeminsights.com> wrote: > Hi Guozhang, > Thanks for the reply. > By taking a lot of time I meant that I see a log message `Restoring state > from changelog topics ....`, followed by just some kafka consumer logs like > `Discovered coordinator....`. Looking at this I assumed that the Stream > threads are waiting for the states to be restored. I am seeing this every > time an instance of the streams application fails and restarts. > > -- > Anish Samir Mashankar > Site Reliability Engineer > VIMANA > +91-9789870733 > -- -- Guozhang