> > > > > > > wrote:
> > > > > > >
> > > > > > > > Hi all,
> > > > > > > >
> > > > > > > > If I don't see additional feedback in the next few days, I'll
> > start a
> > > > > > > vote
gt; Hi all,
> > > > > > > >
> > > > > > > > I've updated the KIP to reflect the latest discussions.
> > > > > > > >
> > > > > > > > Tom,
> > > > > > > > 2) Updated
> > > &
gt; > > > reasons in the Rejected Alternatives section
> > > > > > >
> > > > > > > Please take a look and let me know if you have any feedback.
> > > > > > > Thanks
> > > > > > >
> > > > > >
; > > >
> > > > > > > Thanks for the feedback.
> > > > > > > We also need to assign replicas when adding partitions to an
> > > > > > > existing
> > > > > > > topic. This is why I choose to use a list of partition ids.
>
d the starting partition id.
> > > > > >
> > > > > > Let me know if you have more questions
> > > > > >
> > > > > > On Tue, Oct 6, 2020 at 2:16 AM Efe Gencer
> > >
> > > > wrote:
> > &g
; > > Hi Mickael,
> > > > > >
> > > > > > Thanks for the KIP!
> > > > > > A call to an external system, e.g. Cruise Control, in the
> > > implementation of the provided interface can indeed help with the initial
>
> > > > > A call to an external system, e.g. Cruise Control, in the
> > implementation of the provided interface can indeed help with the initial
> > assignment of partitions.
> > > > >
> > > > > I am curious why the proposed
> > `Re
curious why the proposed
> `ReplicaAssignor#assignReplicasToBrokers` receives a list of partition ids
> as opposed to the number of partitions to create the topic with?
> > > >
> > > > Would you clarify if this API is expected to be used (1) only for
> new topics or (2) also f
tion ids as opposed to the number of partitions
> > > to create the topic with?
> > >
> > > Would you clarify if this API is expected to be used (1) only for new
> > > topics or (2) also for existing topics?
> > >
> > > Best,
> > > Efe
> > > _
c with?
> >
> > Would you clarify if this API is expected to be used (1) only for new
> > topics or (2) also for existing topics?
> >
> > Best,
> > Efe
> > ____
> > From: Mickael Maison
> > Sent: Thursday, October 1, 202
Best,
> Efe
>
> From: Mickael Maison
> Sent: Thursday, October 1, 2020 9:43 AM
> To: dev
> Subject: Re: [DISCUSS] KIP-660: Pluggable ReplicaAssignor
>
> Thanks Tom for the feedback!
>
> 1. If the data returned by the Rep
Subject: Re: [DISCUSS] KIP-660: Pluggable ReplicaAssignor
Thanks Tom for the feedback!
1. If the data returned by the ReplicaAssignor implementation does not
match that was requested, we'll also throw a ReplicaAssignorException
2. Good point, I'll update the KIP
3. The KIP mentions an
Thanks Tom for the feedback!
1. If the data returned by the ReplicaAssignor implementation does not
match that was requested, we'll also throw a ReplicaAssignorException
2. Good point, I'll update the KIP
3. The KIP mentions an error code associated with
ReplicaAssignorException: REPLICA_ASSIGNO
Hi Mickael,
A few thoughts about the ReplicaAssignor contract:
1. What happens if a ReplicaAssignor impl returns a Map where some
assignments don't meet the given replication factor?
2. Fixing the signature of assignReplicasToBrokers() as you have would make
it hard to pass extra information in t
Thanks Mickael, I think adding the new Exception resolves my concerns.
On Thu, Sep 3, 2020 at 9:47 AM Mickael Maison
wrote:
> Thanks Robert and Ryanne for the feedback.
>
> ReplicaAssignor implementations can throw an exception to indicate an
> assignment can't be computed. This is already what
Thanks Robert and Ryanne for the feedback.
ReplicaAssignor implementations can throw an exception to indicate an
assignment can't be computed. This is already what the current round
robin assignor does. Unfortunately at the moment, there are no generic
error codes if it fails, it's either INVALID_
Thanks Mickael, the KIP makes sense to me, esp for cases where an external
system (like cruise control or an operator) knows more about the target
cluster state than the broker does.
Ryanne
On Thu, Aug 20, 2020, 10:46 AM Mickael Maison
wrote:
> Hi,
>
> I've created KIP-660 to make the replica a
Hi Mickael,
Thanks for the KIP!
One question I have is around failure cases. Are ReplicaAssignor
implementations expected to always compute an assignment, or is it possible
for them to have unsatisfiable conditions? One example I can think of is a
requirement that at least one partition be placed
Hi,
I've created KIP-660 to make the replica assignment logic pluggable.
https://cwiki.apache.org/confluence/display/KAFKA/KIP-660%3A+Pluggable+ReplicaAssignor
Please take a look and let me know if you have any feedback.
Thanks
19 matches
Mail list logo
