Apologize for the late comment on this... So fair assignment by count (taking into account the current partition count of each broker) is very good. However, it's worth noting that all partitions are not created equal. We have actually been performing more rebalance work based on the partition size on disk, as given equal retention of all topics, the size on disk is a better indicator of the amount of traffic a partition gets, both in terms of storage and network traffic. Overall, this seems to be a better balance.
In addition to this, I think there is very much a need to have Kafka be rack-aware. That is, to be able to assure that for a given cluster, you never assign all replicas for a given partition in the same rack. This would allow us to guard against maintenances or power failures that affect a full rack of systems (or a given switch). I think it would make sense to implement the reassignment logic as a pluggable component. That way it would be easy to select a scheme when performing a reassignment (count, size, rack aware). Configuring a default scheme for a cluster would allow for the brokers to create new topics and partitions in compliance with the requested policy. -Todd On Thu, Jan 22, 2015 at 10:13 PM, Joe Stein <joe.st...@stealth.ly> wrote: > I will go back through the ticket and code and write more up. Should be > able to-do that sometime next week. The intention was to not replace > existing functionality by issue a WARN on use. The following version it is > released we could then deprecate it... I will fix the KIP for that too. > > On Fri, Jan 23, 2015 at 12:34 AM, Neha Narkhede <n...@confluent.io> wrote: > > > Hey Joe, > > > > 1. Could you add details to the Public Interface section of the KIP? This > > should include the proposed changes to the partition reassignment tool. > > Also, maybe the new option can be named --rebalance instead of > > --re-balance? > > 2. It makes sense to list --decommission-broker as part of this KIP. > > Similarly, shouldn't we also have an --add-broker option? The way I see > > this is that there are several events when a partition reassignment is > > required. Before this functionality is automated on the broker, the tool > > will generate an ideal replica placement for each such event. The users > > should merely have to specify the nature of the event e.g. adding a > broker > > or decommissioning an existing broker or merely rebalancing. > > 3. If I understand the KIP correctly, the upgrade plan for this feature > > includes removing the existing --generate option on the partition > > reassignment tool in 0.8.3 while adding all the new options in the same > > release. Is that correct? > > > > Thanks, > > Neha > > > > On Thu, Jan 22, 2015 at 9:23 PM, Jay Kreps <jay.kr...@gmail.com> wrote: > > > > > Ditto on this one. Can you give the algorithm we want to implement? > > > > > > Also I think in terms of scope this is just proposing to change the > logic > > > in ReassignPartitionsCommand? I think we've had the discussion various > > > times on the mailing list that what people really want is just for > Kafka > > to > > > do it's best to balance data in an online fashion (for some definition > of > > > balance). i.e. if you add a new node partitions would slowly migrate to > > it, > > > and if a node dies, partitions slowly migrate off it. This could > > > potentially be more work, but I'm not sure how much more. Has anyone > > > thought about how to do it? > > > > > > -Jay > > > > > > On Wed, Jan 21, 2015 at 10:11 PM, Joe Stein <joe.st...@stealth.ly> > > wrote: > > > > > > > Posted a KIP for --re-balance for partition assignment in > reassignment > > > > tool. > > > > > > > > > > > > > > > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-6+-+New+reassignment+partition+logic+for+re-balancing > > > > > > > > JIRA https://issues.apache.org/jira/browse/KAFKA-1792 > > > > > > > > While going through the KIP I thought of one thing from the JIRA that > > we > > > > should change. We should preserve --generate to be existing > > functionality > > > > for the next release it is in. If folks want to use --re-balance then > > > > great, it just won't break any upgrade paths, yet. > > > > > > > > /******************************************* > > > > Joe Stein > > > > Founder, Principal Consultant > > > > Big Data Open Source Security LLC > > > > http://www.stealth.ly > > > > Twitter: @allthingshadoop <http://www.twitter.com/allthingshadoop> > > > > ********************************************/ > > > > > > > > > > > > > > > -- > > Thanks, > > Neha > > >
