rebalancing leadership quite often

[Sa Li <sa...@gmail.com>]

Hi, All

My dev cluster has three nodes (1, 2, 3), but I've seen quite often that
the 1 node just not work as a leader, I run preferred-replica-election many
time, every time I run replica election, I see 1 turn out to be leader for
some partitions, but it just stop leadership after a while, and the that
partitions transfer to other partitions, meaning it will not be used for
consumers.

I thought that happens only that broker 1 was crashed or stopped, but it
didn't, but I still see that leadership shift. Any ideas about this?

thanks

-- 

Alec Li

Re: rebalancing leadership quite often

[Joe Stein <jo...@stealth.ly>]

https://cwiki.apache.org/confluence/display/KAFKA/FAQ#FAQ-Whypartitionleadersmigratethemselvessometimes
?

~ Joe Stein
- - - - - - - - - - - - - - - - -

  http://www.stealth.ly
- - - - - - - - - - - - - - - - -

On Fri, Feb 20, 2015 at 7:19 PM, Sa Li <sa...@gmail.com> wrote:

> Hi, All
>
> My dev cluster has three nodes (1, 2, 3), but I've seen quite often that
> the 1 node just not work as a leader, I run preferred-replica-election many
> time, every time I run replica election, I see 1 turn out to be leader for
> some partitions, but it just stop leadership after a while, and the that
> partitions transfer to other partitions, meaning it will not be used for
> consumers.
>
> I thought that happens only that broker 1 was crashed or stopped, but it
> didn't, but I still see that leadership shift. Any ideas about this?
>
> thanks
>
> --
>
> Alec Li
>