Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
Kafka >> mail # user >> hadoop-consumer code in contrib package


+
navneet sharma 2013-01-14, 18:35
+
Felix GV 2013-01-14, 22:43
+
navneet sharma 2013-01-15, 17:06
+
Felix GV 2013-01-15, 18:17
+
navneet sharma 2013-01-17, 00:41
+
Jun Rao 2013-01-17, 05:12
+
navneet sharma 2013-01-17, 14:21
Copy link to this message
-
Re: hadoop-consumer code in contrib package
That may be an alternative feasible approach. You can
call ConsumerConnector.shutdown() to close the consumer cleanly.

Thanks,

Jun

On Thu, Jan 17, 2013 at 6:20 AM, navneet sharma <[EMAIL PROTECTED]
> wrote:

> That makes sense.
>
> I tried an alternate approach- i am using high level consumer and going
> through Hadoop HDFS APIs and pushing data in HDFS.
>
> I am not creating any jobs for that.
>
> The only problem i am seeing here is that the consumer is designed to run
> forever. Which means i need to find out how to close the HDFS file and kill
> consumer.
>
> Is there any way to kill or close high level consumer gracefully?
>
> I am running v0.7.0. I don't mind upgrading to higher version if that
> allows me this kind of consumer handling.
>
> Thanks,
> Navneet
>
>
> On Thu, Jan 17, 2013 at 10:41 AM, Jun Rao <[EMAIL PROTECTED]> wrote:
>
> > I think the main reason for using SimpleConsumer is to manage offsets
> > explicitly. For example, this is useful when Hadoop retries failed tasks.
> > Another reason is that Hadoop already does load balancing. So, there is
> not
> > much need to balance the load again using the high level consumer.
> >
> > Thanks,
> >
> > Jun
> >
> > On Wed, Jan 16, 2013 at 4:40 PM, navneet sharma <
> > [EMAIL PROTECTED]
> > > wrote:
> >
> > > Thanks Felix.
> > >
> > > One question still remains. Why SimpleConsumer?
> > > Why not high level Consumer? If i change the code to high level
> consumer,
> > > will it create any challenges?
> > >
> > >
> > > Navneet
> > >
> > >
> > > On Tue, Jan 15, 2013 at 11:46 PM, Felix GV <[EMAIL PROTECTED]> wrote:
> > >
> > > > Please read the Kafka design paper <
> > http://kafka.apache.org/design.html
> > > >.
> > > >
> > > > It may look a little long, but it's as short as it can be. Kafka
> > differs
> > > > from other messaging system in a couple of ways, and it's important
> to
> > > > understand the fundamental design choices that were made in order to
> > > > understand the way Kafka works.
> > > >
> > > > I believe my previous email already answers both your offset tracking
> > and
> > > > retention questions, but if my explanation are not clear enough, then
> > the
> > > > next best thing is probably to read the design paper :)
> > > >
> > > > --
> > > > Felix
> > > >
> > > >
> > > > On Tue, Jan 15, 2013 at 12:01 PM, navneet sharma <
> > > > [EMAIL PROTECTED]> wrote:
> > > >
> > > > > Thanks Felix for sharing your work. Contrib hadoop-consumer looks
> > like
> > > > the
> > > > > same way.
> > > > >
> > > > > I think i need to really understand this offset stuff. So far i
> have
> > > used
> > > > > only high level consumer.When consumer is done reading all the
> > > messages,
> > > > i
> > > > > used to kill the process(because it won't on its own).
> > > > >
> > > > > Again i used Producer to pump more messages and Consumer to read
> the
> > > new
> > > > > messages(which is a new process as i killed the last consumer).
> > > > >
> > > > > But i never saw messages getting duplicating.
> > > > >
> > > > > Now its not very clear for me that how offsets is tracked
> > specifically
> > > > when
> > > > > i am re-launching the consumer?
> > > > > And why retention policy is not working when used with
> > SimpleConsumer?
> > > > For
> > > > > my experiment i made it 4 hours.
> > > > >
> > > > > Please help me understand.
> > > > >
> > > > > Thanks,
> > > > > Navneet
> > > > >
> > > > >
> > > > > On Tue, Jan 15, 2013 at 4:12 AM, Felix GV <[EMAIL PROTECTED]>
> > wrote:
> > > > >
> > > > > > I think you may be misunderstanding the way Kafka works.
> > > > > >
> > > > > > A kafka broker is never supposed to clear messages just because a
> > > > > consumer
> > > > > > read them.
> > > > > >
> > > > > > The kafka broker will instead clear messages after their
> retention
> > > > period
> > > > > > ends, though it will not delete the messages at the exact time
> when
> > > > they
> > > > > > expire. Instead, a background process will periodically delete a