Hello,

1. KTable is an API class that only exists in the Kafka Streams DSL, in
which users cannot directly query it by key, but can use operators like
joins and aggregates that will query the KTable internal for the operation.

2. When involved in stateful operations like joins, KTable is usually
implemented by "materializing" the corresponding stream into persistent
storage, which is rocksDB today.

3. The rocksDB store class does provide "put / get / etc" APIs for by-key
queries, which is used in the DSL implementation for joins, for example.
User can also user these interfaces with the Processor API, in which they
can define an arbitrary "state store", and access that store with the
provided functions.


Guozhang

On Fri, May 6, 2016 at 10:31 AM, Spico Florin <spicoflo...@gmail.com> wrote:

> hi!
>  thank you very much for your answer. what is unclear for me if can query
> the ktable using the keys and what is the relation between ktable and
> rocksdb?
>  i look forward for your answers.
> regards,
>  florin
>
> On Friday, May 6, 2016, Guozhang Wang <wangg...@gmail.com> wrote:
>
> > Hello,
> >
> > KTable's changelog is using log compaction by default instead of log
> > deletion. But we are adding mechanism to to let users configure their
> > changelog topics.
> >
> > Guozhang
> >
> >
> > On Thu, May 5, 2016 at 12:00 PM, Spico Florin <spicoflo...@gmail.com
> > <javascript:;>> wrote:
> >
> > > hello!
> > >  i would like to ask you if ktable is using as backend  storage a
> > compacted
> > > topic ? i have read here
> > >
> > >
> >
> http://www.confluent.io/blog/introducing-kafka-streams-stream-processing-made-simple
> > > thtat a ktable is at its base a compacted topic
> > >
> > > if i would like to have the messages from ktable be persitent forever (
> > do
> > > no delete them after the 7 days as the default deletion policy) what
> > should
> > > i take into consideration in order do not mess up the compaction
> process
> > > that shoul remove the duplicates?
> > > i look forward for your answers.
> > > regards,
> > >  florin
> > >
> >
> >
> >
> > --
> > -- Guozhang
> >
>



-- 
-- Guozhang

Reply via email to