Thanks for the KIP, +1 (binding) with a couple of minor suggestions:
//Sample configuration for throttled replicas
{
"version":1,
"config": {
"throttled-replicas":"0-0:0-1:0-2:1-0:1-1:1-2"
}
}
I think it would be nicer if the "throttled-replicas" value was a JSON
array instead of a String. So:
//Sample configuration for throttled replicas
{
"version":1,
"config": {
"throttled-replicas":["0-0","0-1","0-2","1-0","1-1","1-2"]
}
}
It does take a little more space, but it's more standard. Do we think
the space savings are worth
coming up with our own encoding?
And about SumReplicaLag, I think ReplicaLagSum sounds a bit better.
Obviously subjective so
if you prefer the current name feel free to keep it.
Ismael
On Mon, Aug 22, 2016 at 11:58 PM, Jun Rao <j...@confluent.io> wrote:
> Ben,
>
> Thanks for the proposal. +1.
>
> Just a few minor comments below.
>
> 1. We have a LeaderOverThrottledRate metric to indicate the amount of
> throttling happening in the leader broker. It seems that we should have the
> equivalent of that for the follower to indicate the amount of throttling in
> the follower, if any.
> 2. Do we still need the PartitionBytesInRate metric? There is no reference
> on how it's going to be used.
> 3. In the test plan, you mentioned "Then replicas should move at close to
> (but no more than) than (the quota dictated rate - the inbound rate).". It
> seems that the replicas should always be moved at the quota rate
> independent whether there is incoming traffic from the producer?
>
> Jun
>
> On Fri, Aug 19, 2016 at 1:21 AM, Ben Stopford <b...@confluent.io> wrote:
>
> > I’d like to initiate the voting process for KIP-73:
> > https://cwiki.apache.org/confluence/display/KAFKA/KIP-
> > 73+Replication+Quotas <https://cwiki.apache.org/
> > confluence/display/KAFKA/KIP-73+Replication+Quotas>
> >
> > Ben
>
