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

Switch to Threaded View
Kafka, mail # user - Questions on kafka leader selection


Copy link to this message
-
Re: Questions on kafka leader selection
Jun Rao 2013-06-20, 05:19
We do have a tool ReassignPartitionsCommand that can move existing replicas
to new brokers. It hasn't been thoroughly tested though.

Thanks,

Jun
On Wed, Jun 19, 2013 at 9:41 PM, Yogesh Sangvikar <
[EMAIL PROTECTED]> wrote:

> Hi Jun,
>
> Thanks for the reply.
>
> But this case arises one more questions that, If all the 3 down brokers are
> crashed and we have added
> some new brokers, is there any way to re-balance those replicas to make
> partition available? OR any tool to balance the number of replicas if any
> one of the ISR goes down but the controller active? As this helps to make
> high data availability.
>
> Thanks,
>
> Yogesh Sangvikar
>
>
> On Wed, Jun 19, 2013 at 8:53 PM, Jun Rao <[EMAIL PROTECTED]> wrote:
>
> > If leader is -1, that indicates the partition is unavailable. Were
> replicas
> > 4,2,3 all down at that time? If so, you need to start at least one of
> those
> > brokers to make the partition available again.
> >
> > Thanks,
> >
> > Jun
> >
> >
> > On Wed, Jun 19, 2013 at 5:05 AM, Yogesh Sangvikar <
> > [EMAIL PROTECTED]> wrote:
> >
> > > Hi Team,
> > >
> > > Please help to understand following questing as we are planning to use
> > > kafka 0.8 in our production,
> > > 1. We found the 'leader: -1' state, when the leader selection will
> happen
> > > for such partitions? Does the partition data will be unavailable till
> > that
> > > time?
> > >
> > > example:
> > > topic: test1    partition: 0    leader: -1      replicas: 4,2,3 isr:
> > > topic: test1    partition: 1    leader: 0       replicas: 0,3,4 isr:
> 0,4
> > > topic: test1    partition: 2    leader: 1       replicas: 1,4,0 isr:
> > 1,0,4
> > >
> > > 2. Do we need to restart the brokers to update the isr entries to
> remove
> > > the unavailable brokers?
> > >
> > > Thanks,
> > > Yogesh Sangvikar
> > >
> >
>