[ https://issues.apache.org/jira/browse/KAFKA-16344?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17824385#comment-17824385 ]
Janardhana Gopalachar edited comment on KAFKA-16344 at 3/7/24 2:05 PM: ----------------------------------------------------------------------- HI [~gharris1727] Currently in our mirror maker spec we have offset.lag.max: 0, so should it be set to max 100, Will this reduce the through put on source topics or target topic, basically couldnt find much documentation for the same. Is the offset.lag.max value set to 0 is contributing for CPU load ? what would be value that could be set if it to be more than 100? {code:java} sourceConnector: maxTasks: 12 settings: consumer.max.poll.records: 2000 consumer.ssl.endpoint.identification.algorithm: "" offset-syncs.topic.location: target offset-syncs.topic.replication.factor: 3 offset.lag.max: 0 producer.ssl.endpoint.identification.algorithm: "" refresh.topics.interval.seconds: 300 replication.factor: 3 replication.policy.class: org.apache.kafka.connect.mirror.IdentityReplicationPolicy replication.policy.separator: "" ssl.endpoint.identification.algorithm: "" sync.topic.acls.enabled: "false"{code} and along with we have below questions 1. What is the proportion of mm2-offsetsyncsinternal topic writes to overall MM2 traffic? 2. Is there a way to tune the internal topic writes with increased traffic MM2? 3. I want to understand the expected mm2-offsetsyncsinternal write TPS for a given amount of MM2 traffic. Are there any tunable parameters to reduce these writes, and what are the consequences of tuning if any? 4. In a larger Kafka cluster, if a single broker is overloaded with mm2-offsetsyncsinternal traffic, can it lead to a broker crash? Are there any guidelines available for such scenarios? Currently, we have 6 brokers with 24K MM2 traffic, and internal writes are at 10K, resulting in a 20% CPU increase on one broker. 5. Are there any limitations on Kafka brokers scaling , I mean how much kafka broker can be expanded in single Kakfa cluster due to the mm2-offsetsyncsinternal topic? 6. How can the system be dimensioned to handle MM2 internal topic writes effectively? Are there any recommended figures available? For instance, for a given amount of traffic (X), what percentage increase in CPU (Y) should each broker have to handle MM2 internal topic writes? Note that in other pods, this resource may not be utilized. Regards Jana was (Author: JIRAUSER301625): HI [~gharris1727] Below are few questions we have 1. What is the proportion of mm2-offsetsyncsinternal topic writes to overall MM2 traffic? 2. Is there a way to tune the internal topic writes with increased traffic MM2? 3. I want to understand the expected mm2-offsetsyncsinternal write TPS for a given amount of MM2 traffic. Are there any tunable parameters to reduce these writes, and what are the consequences of tuning if any? 4. In a larger Kafka cluster, if a single broker is overloaded with mm2-offsetsyncsinternal traffic, can it lead to a broker crash? Are there any guidelines available for such scenarios? Currently, we have 6 brokers with 24K MM2 traffic, and internal writes are at 10K, resulting in a 20% CPU increase on one broker. 5. Are there any limitations on Kafka brokers scaling , I mean how much kafka broker can be expanded in single Kakfa cluster due to the mm2-offsetsyncsinternal topic? 6. How can the system be dimensioned to handle MM2 internal topic writes effectively? Are there any recommended figures available? For instance, for a given amount of traffic (X), what percentage increase in CPU (Y) should each broker have to handle MM2 internal topic writes? Note that in other pods, this resource may not be utilized. Regards Jana > Internal topic mm2-offset-syncs<clustername>internal created with single > partition is putting more load on the broker > --------------------------------------------------------------------------------------------------------------------- > > Key: KAFKA-16344 > URL: https://issues.apache.org/jira/browse/KAFKA-16344 > Project: Kafka > Issue Type: Bug > Components: connect > Affects Versions: 3.5.1 > Reporter: Janardhana Gopalachar > Priority: Major > > We are using Kafka 3.5.1 version, we see that the internal topic created by > mirrormaker > mm2-offset-syncs<clustername>internal is created with single partition due to > which the CPU load on the broker which will be leader for this partition is > increased compared to other brokers. Can multiple partitions be created for > the topic so that the CPU load would get distributed > > Topic: mm2-offset-syncscluster-ainternal TopicId: XRvTDbogT8ytNhqX2YTyrA > PartitionCount: 1ReplicationFactor: 3 Configs: > min.insync.replicas=2,cleanup.policy=compact,message.format.version=3.0-IV1 > Topic: mm2-offset-syncscluster-ainternal Partition: 0 Leader: 2 > Replicas: 2,1,0 Isr: 2,1,0 -- This message was sent by Atlassian Jira (v8.20.10#820010)