Hi team, few questions on Kafka supervisor behaviour. If anyone can provide pointers on this it will be helpful. Thanks: 1. With

stopTaskCount=2

,

taskCount=4

and

replicas=2

, does

stopTaskCount

stop 2 task groups (all replicas included), or 2 individual tasks(staggering the two replicas)? 2. With same config as above, will both replicas of a task group start at the same time? During

intermediatePersistPeriod

flushes, will both hit persist roughly simultaneously? If one peon is blocked on merge buffer acquisition, can the

broker

use the other replica's peon to serve the query, or does it wait for both?

1- I believe

stopTaskCount

is an entire group, so

2

means 2 task groups

thanks @Gian Merlino for checking Is there way we can ensure task replica are not in sync explicitly, so that their flushes during

intermediatePersistPeriod

and

handoff

don't happen at same time. If we could offset them, the broker could route to the non-persisting replica when one is blocked on merge buffer acquisition during persist/handoff

what about persist makes it be blocked on merge buffer acquisition? I haven't noticed that before

@Gian Merlino currently we have

intermediatePersistPeriod

set to 5m and we notice every 5m, the query timeouts as there seems to be resource contention as most of thread/buffer are utilised by ingestion + persist, leaving lesser for query to utilise when running

hmm. the thread and buffer pools used by ingest and query are not shared with each other. It could be that the task is running in an overall resource-constrained or oversubscribed situation (VM, container, etc)

oh ok, thanks

If too many realtime queries are coming to a single hot MiddleManager node, would having two replicas help distribute that load across two peons on two different nodes?

Yes that would help. Task replicas will bind to different MMs and the broker will load balance queries across them. By default,

druid.broker.balancer.type

(defaults to random) determines how the broker will balance queries to historicals & realtime tasks: https://druid.apache.org/docs/latest/configuration/#query-configuration

ack, thanks for details

Yep, I think we could rephrase that sentence 🙂 Do you want to contribute?

sure