# WARN kafka message: client/metadata found some partitions to be leaderless **URL:** \ **Category:** Beats\ **Tags:** filebeat\ **Created:** [December 27, 2017, 9:57am UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293 "2017-12-27T09:57:47Z")\ **Posts on this page:** 5\ **Page:** 1
**Author:** ![Abhijit\_Paul](https://avatars.discourse-cdn.com/v4/letter/a/8e8cbc/32.png) [@Abhijit\_Paul](https://discuss.elastic.co/u/Abhijit_Paul)\ **Post date:** [December 27, 2017, 9:57am UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293/1 "2017-12-27T09:57:47Z")
Hi All, I am unable to post a message from filebeat as an external producer (outside k8s cluster) to kafka-inside kubernetes cluster. The topic "test\_topic" is already created. Error from filebeat: > 2017/12/27 09:26:03.225021 log.go:16: WARN kafka message: Initializing new client > 2017/12/27 09:26:03.225076 log.go:12: WARN client/metadata fetching metadata for all topics from broker 10.19.166.10:30094 > { > "@timestamp": "2017-12-27T09:26:00.224Z", > "beat": { > "hostname": "", > "name": "", > "version": "5.5.1" > }, > "input\_type": "log", > "message": "test message", > "offset": 15, > "source": "/root/fb/input\_logs/art.log", > "type": "log" > } > 2017/12/27 09:26:03.264764 log.go:12: WARN Connected to broker at 10.19.166.10:30094 (unregistered) > 2017/12/27 09:26:03.265964 log.go:12: WARN client/brokers registered new broker #0 at 192.168.1.13:9094 > 2017/12/27 09:26:03.266037 log.go:16: WARN kafka message: Successfully initialized new client > 2017/12/27 09:26:03.266198 log.go:12: WARN client/metadata fetching metadata for [test\_topic] from broker 10.19.166.10:30094 > 2017/12/27 09:26:03.266995 log.go:16: WARN kafka message: client/metadata found some partitions to be leaderless > 2017/12/27 09:26:03.267009 log.go:12: WARN client/metadata retrying after 250ms... (3 attempts remaining) Debugging steps followed: Step1: > **kubectl exec kafka-0 -n elk -c kafka curl 10.19.166.10:30094** > % Total % Received % Xferd Average Speed Time Time Time Current > Dload Upload Total Spent Left Speed > 100 14 0 14 0 0 3879 0 --:--:-- --:--:-- --:--:-- 4666 Step2: > **kubectl exec kafka-0 -n elk -c kafka curl 192.168.1.13:9094** > % Total % Received % Xferd Average Speed Time Time Time Current > Dload Upload Total Spent Left Speed > 100 14 0 14 0 0 9575 0 --:--:-- --:--:-- --:--:-- 14000 Step3: kafka - **server.properties** > listeners=SASL\_PLAINTEXT://192.168.1.13:9092,PLAINTEXT://192.168.1.13:9093,SSL://192.168.1.13:9094 Step4: > **kubectl get ep -n elk** > kafka-headless-external 192.168.1.13:9094 > **kubectl get svc -n elk -o wide** > kafka-headless-external 172.31.38.80 9094:30094/TCP ---
**Author:** ![steffens](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/steffens/32/79630_2.png) [@steffens](https://discuss.elastic.co/u/steffens)\ **Post date:** [January 4, 2018, 12:35pm UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293/2 "2018-01-04T12:35:02Z")
Sounds more like a problem with your kafka setup, not beats. In kafka topics are split into partitions. Normally each broker handles/stores a subset of partitions for a given topic. That is, each partition requires one active 'leader' and multiple 'replica' broker. The leader is responsible handling events from publisher and for updating the replicas. Every now and then a new leader is selected. If no leader is available, no events can be processed for said partition. By default the kafka output blocks if some partition can not be handled. This can be disabled by setting `partition.round_robin.reachable_only: false` in the kafka output (see sample in docs: [https://www.elastic.co/guide/en/beats/filebeat/current/kafka-output.html#kafka-output](https://www.elastic.co/guide/en/beats/filebeat/current/kafka-output.html#kafka-output)). It's not really recommende to disable this setting, as in worst case all events might be published to one partition only -\> no more load balancing on publisher, disk and consumers (as only one consumer can read from a partition). ---
**Author:** ![Abhijit\_Paul](https://avatars.discourse-cdn.com/v4/letter/a/8e8cbc/32.png) [@Abhijit\_Paul](https://discuss.elastic.co/u/Abhijit_Paul)\ **Post date:** [January 5, 2018, 9:20am UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293/3 "2018-01-05T09:20:15Z")
Thanks for the explanination....Actually that problem is solved now with the change by altering topic partition count from 3 to 1, Not sure with default partitioncount of 3 why it's not worked. ---
**Author:** ![steffens](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/steffens/32/79630_2.png) [@steffens](https://discuss.elastic.co/u/steffens)\ **Post date:** [January 5, 2018, 11:59am UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293/4 "2018-01-05T11:59:53Z")
Uhm... you still have replication enabled? Are your kafka brokers correctly interconnected to form a cluster? ---
**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\ **Post date:** [February 2, 2018, 12:00pm UTC](https://discuss.elastic.co/t/warn-kafka-message-client-metadata-found-some-partitions-to-be-leaderless/113293/5 "2018-02-02T12:00:10Z")
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.