> > For reference, this is the label that we have in our consumer pod: > > client.rack=us-east-1a > > So we are assuming that by having this label in the pod, this consumer pod > will fetch data from the broker present in us-east-1a AZ. It would be > helpful to know if we have to set this somewhere inside the pod as an > environment variable as well. >
Do you just add *client.rack* as a label or also expose it as a consumer config somehow to the actual java client? Unless you have some custom entrypoint in your image, which reads labels and pass them as a configs, this won't work. On Fri, Nov 1, 2024 at 2:22 PM Soham Chakraborty <dec.so...@gmail.com> wrote: > And kafka client version is 2.8.0 as well. I realized that I didn't answer > in last message. > > Thanks, > > On Fri, Nov 1, 2024 at 2:52 PM Soham Chakraborty <dec.so...@gmail.com> > wrote: > > > Hi Artem, > > > > Yes, everything is in the same AWS account and same region. The goal is > to > > minimize the intra-AZ data transfer cost. > > > > On Fri, Nov 1, 2024 at 2:31 PM Artem Timchenko > > <artem.timche...@bolt.eu.invalid> wrote: > > > >> > > >> > For reference, this is the label that we have in our consumer pod: > >> > > >> > client.rack=us-east-1a > >> > > >> > >> Are your clients in the same AWS account as your Kafka cluster is? > >> > >> On Fri, Nov 1, 2024 at 10:46 AM Soham Chakraborty <dec.so...@gmail.com> > >> wrote: > >> > >> > Hi, > >> > > >> > Thank you for replying. > >> > > >> > We are using Java as the client library. > >> > This is the version: > >> > > >> > $ java -version > >> > openjdk version "14.0.1" 2020-04-14 > >> > OpenJDK Runtime Environment (build 14.0.1+7) > >> > OpenJDK 64-Bit Server VM (build 14.0.1+7, mixed mode, sharing) > >> > > >> > Is there any article where I can find out whether this Java version is > >> KIP > >> > 392 compatible? > >> > > >> > For reference, this is the label that we have in our consumer pod: > >> > > >> > client.rack=us-east-1a > >> > > >> > So we are assuming that by having this label in the pod, this consumer > >> pod > >> > will fetch data from the broker present in us-east-1a AZ. It would be > >> > helpful to know if we have to set this somewhere inside the pod as an > >> > environment variable as well. > >> > > >> > Thanks, > >> > > >> > > >> > On Wed, Oct 30, 2024 at 6:47 PM Ömer Şiar Baysal < > osiarbay...@gmail.com > >> > > >> > wrote: > >> > > >> > > Hi, > >> > > > >> > > The log shown on the blog post was for the console consumer tool. > >> These > >> > > log messages emitted for the consumer process so you need to > configure > >> > > log4j configuration for the kafka tools if you want to achieve the > >> same > >> > > thing. > >> > > > >> > > The functionality provided by the KIP-392 is only realized only if > >> client > >> > > implementation supports it. So the question is which library are you > >> > using > >> > > for the consumers? > >> > > > >> > > Regards, > >> > > OSB > >> > > > >> > > On Wed, Oct 30, 2024, 13:51 Soham Chakraborty <dec.so...@gmail.com> > >> > wrote: > >> > > > >> > > > Hi, > >> > > > > >> > > > We are trying to implement the feature from KIP 392 to allow > >> consumers > >> > to > >> > > > fetch from the closest replica. We are using AWS and Kubernetes > >> (kOps) > >> > > and > >> > > > the consumers do not always fetch from the broker in the same AZ, > >> > > resulting > >> > > > in cross AZ traffic cost. The entire purpose of this mail/exercise > >> is > >> > to > >> > > > solve that. > >> > > > > >> > > > There is a very nice documentation from AWS which tells what and > >> how to > >> > > do: > >> > > > > >> > > > > >> > > > >> > > >> > https://aws.amazon.com/blogs/big-data/reduce-network-traffic-costs-of-your-amazon-msk-consumers-with-rack-awareness/ > >> > > > > >> > > > We have followed all these steps: > >> > > > > >> > > > * added the two configuration parameters on kafka brokers (EC2 > >> > instances) > >> > > > * added a label `client.rack=<AZ_WHERE_THE_K8S_POD_IS_RUNNING>` to > >> the > >> > > pod > >> > > > with an admission controller (kyverno - which mutates the pods > >> during > >> > > > placement to nodes) > >> > > > > >> > > > After doing these, we expected the traffic cost to go down but AWS > >> cost > >> > > > explorer thinks otherwise, which is why we are thinking that > >> something > >> > > > might be wrong somewhere. To find out we are resorting to good ol > >> logs > >> > > from > >> > > > kafka brokers. And that is the reason for this mail, we haven't > been > >> > able > >> > > > to come up with a good logging configuration that shows us from > >> where > >> > the > >> > > > consumers are fetching data. We are expecting to see something > like > >> > what > >> > > > the blog article from AWS is saying: > >> > > > > >> > > > +++++ > >> > > > [2022-04-27 18:04:18,200] DEBUG [Consumer > >> > > > clientId=consumer-console-consumer-99846-1, > >> > > groupId=console-consumer-99846] > >> > > > Added READ_UNCOMMITTED fetch request for partition order-0 at > >> position > >> > > > FetchPosition{offset=39, offsetEpoch=Optional[0], > >> > > > currentLeader=LeaderAndEpoch{leader=Optional[ > >> > > > b-1.mskcluster-msk.jcojml.c23.kafka.us-east-1.amazonaws.com:9092 > >> (id: > >> > 1 > >> > > > rack: use1-az2)], epoch=0}} to node > >> > > > b-3.mskcluster-msk.jcojml.c23.kafka.us-east-1.amazonaws.com:9092 > >> (id: > >> > 3 > >> > > > rack: use1-az1) > >> (org.apache.kafka.clients.consumer.internals.Fetcher) > >> > > > +++++ > >> > > > > >> > > > We tried using the configuration provided in the article but that > >> also > >> > > > didn't help out. > >> > > > > >> > > > This is our configuration: > >> > > > > >> > > > +++++ > >> > > > # ./kafka-topics.sh --version > >> > > > 2.8.0 (Commit:ebb1d6e21cc92130) > >> > > > > >> > > > # grep -v "^#" ../config/server.properties | grep . > >> > > > broker.id=11 > >> > > > broker.rack=us-east-1a > >> > > > num.network.threads=64 > >> > > > num.io.threads=8 > >> > > > socket.send.buffer.bytes=102400 > >> > > > socket.receive.buffer.bytes=102400 > >> > > > socket.request.max.bytes=104857600 > >> > > > message.max.bytes=10485760 > >> > > > log.dirs=/mnt/data/kafka > >> > > > num.partitions=1 > >> > > > num.recovery.threads.per.data.dir=4 > >> > > > offsets.topic.replication.factor=2 > >> > > > transaction.state.log.replication.factor=2 > >> > > > transaction.state.log.min.isr=1 > >> > > > delete.topic.enable=true > >> > > > auto.create.topics.enable=false > >> > > > log.retention.hours=24 > >> > > > log.segment.bytes=1073741824 > >> > > > log.retention.check.interval.ms=300000 > >> > > > zookeeper.connect=zk11.foo.bar.com,zk12.foo.bar.com, > >> zk13.foo.bar.com, > >> > > > zk14.foo.bar.com,zk15.foo.bar.com > >> > > > zookeeper.connection.timeout.ms=6000 > >> > > > group.initial.rebalance.delay.ms=5000 > >> > > > replica.selector.class = > >> > > > org.apache.kafka.common.replica.RackAwareReplicaSelector > >> > > > > >> > > > # cat ../config/log4j.properties > >> > > > log4j.rootLogger=INFO,ROLLINGFILE > >> > > > > >> > > > log4j.appender.CONSOLE=org.apache.log4j.ConsoleAppender > >> > > > log4j.appender.CONSOLE.Threshold=WARN > >> > > > log4j.appender.CONSOLE.layout=org.apache.log4j.PatternLayout > >> > > > log4j.appender.CONSOLE.layout.ConversionPattern=%c{35}:%L - %m%n > >> > > > > >> > > > log4j.appender.ROLLINGFILE=org.apache.log4j.RollingFileAppender > >> > > > log4j.appender.ROLLINGFILE.Threshold=INFO > >> > > > log4j.appender.ROLLINGFILE.File=/var/log/kafka/server.log > >> > > > log4j.appender.ROLLINGFILE.MaxFileSize=200MB > >> > > > log4j.appender.ROLLINGFILE.MaxBackupIndex=10 > >> > > > log4j.appender.ROLLINGFILE.layout=org.apache.log4j.PatternLayout > >> > > > log4j.appender.ROLLINGFILE.layout.ConversionPattern=%d %p [%t] > >> > %c{35}:%L > >> > > - > >> > > > %m%n > >> > > > > >> > > > > >> log4j.logger.org.apache.kafka.clients.consumer.internals.Fetcher=DEBUG > >> > > > +++++ > >> > > > > >> > > > Can anyone please tell how we should customize the logging > >> properties > >> > to > >> > > > see messages shown above. > >> > > > > >> > > > Thanks, > >> > > > > >> > > > >> > > >> > > >
