Re: Broker error: "failed due to Leader not local for partition"

[Jason Rosenberg <jb...@squareup.com>]

"failed due to leader not local for partition" usually occurs in response
to client requests that make a fetch or produce request to a partition, to
the wrong broker (e.g. to a follower and not the leader for the
partition).  Clients need to make a meta-data request first to determine
the leader replica for a partition, etc.  If you clients that are not well
behaved (e.g. they should re-poll the partition meta-data on error), then
you get these exceptions repeatedly.  However, it's normal to see these
messages occasionally, as it's the primary way that clients get notified of
a partition leadership change.

Jason

On Fri, May 29, 2015 at 3:59 PM, Casey Sybrandy - US <cs...@caci.com>
wrote:

> Hello,
>
>
> For some reason, I'm getting this "failed due to Leader not local for
> partition" on my brokers.  I can't send the logs, but I'm going to try to
> describe what I see.  First, note that this is a fresh set of brokers with
> no data.
>
>
> I do see a series of messages stating "Expanding ISR for partition", so it
> looks like all of the partitions have enough replicas.  I set up two
> partitions for each topic and three replicas each.
>
>
> I see several "Removed fetcher for partitions" and "Added fetcher for
> partitions List" messages, which looks right as it removes then adds the
> partition.  However, the first one I see is a pair of those messages with
> no partitions listed.  Right after them I see a "Removed fetcher..."
> message with all of the topics/partitions listed, but no corresponding
> "Added fetcher" message.
>
>
> That's about it that I see right now.  I know there was a bug a couple
> years ago in an early 0.8 release, but I'm running 0.8.2, so I shouldn't
> have that issue.  Googleing just didn't seem to get me any useful
> information.
>
>
> Thanks.
>
>
> Casey
>