This post was updated on .
Hi,
In the last week I have correctly deployed a flink program which get data from a kafka broker on my local machine. Now I'm trying to produce the same thing but moving the kafka broker on a cluster. I didn't change any line of code, I report it here: DataStream<Tuple2<String,JSONLDObject>> stream = env .addSource(new FlinkKafkaConsumer010<>(TOPIC, new CustomDeserializer(), properties)) .assignTimestampsAndWatermarks(new CustomTimestampExtractor()) .keyBy(0); While I have changed just the Kafka Ip. Data model obviously is not changed. Unfortunately now when I start Flink program I get this: INFO org.apache.kafka.common.utils.AppInfoParser - Kafka version : 0.10.0.1 12:30:48,446 INFO org.apache.kafka.common.utils.AppInfoParser - Kafka commitId : a7a17cdec9eaa6c5 12:30:48,625 INFO org.apache.kafka.clients.consumer.internals.AbstractCoordinator - Discovered coordinator giordano-1-4-200:9092 (id: 2147483647 rack: null) for group groupId. 12:30:48,626 INFO org.apache.kafka.clients.consumer.internals.AbstractCoordinator - Marking the coordinator giordano-1-4-200:9092 (id: 2147483647 rack: null) dead for group groupId I bolded the line that worry me. Then, no data are retrieved buy Kafka although flink continue to perform checkpointing etc normally... Any ideas? P.S: I want to add that this issue occurs either with partitioned topic and non-partitioned. Moreover, when flank and kafka are installed locally on same machine it doesn't occurs. |
Hi, AFAIK, Kafka group coordinators are supposed to always be marked dead, because we use static assignment internally and therefore Kafka's group coordination functionality is disabled. Cheers, Gordon
On 11 August 2017 at 6:43:51 PM, AndreaKinn ([hidden email]) wrote:
|
the kafka version I use is the latest (0.11.0.0). But to be honestly, also locally I use 0.11.0.0 and in that case it works correctly. Anyway the last kafka connector on flink is designed for kafka 0.10.x.x
I use OS X locally and Ubuntu on the cluster. It has importance? |
No, there should be no difference between setting it up on Ubuntu or OS X. I can’t really tell any anything suspicious from the information provided so far, unfortunately. Perhaps you can try first checking that the Kafka topic is consumable from where you’re running Flink, e.g. using the example console consumer / producers? On 11 August 2017 at 7:06:46 PM, AndreaKinn ([hidden email]) wrote:
|
I tried running console consumer-producer from the localhost on the cluster: this say me that the broker is currently active.
To reach the cluster from outside I use a redirect from a public (ip, port), because the ip of the kafka broker is private... I suspect the problem can be there. |
I just tried to use telnet to public ip:port from outside and it works.
|
In reply to this post by Tzu-Li (Gordon) Tai
Hi,
You mentioned that your kafka broker is behind a proxy. This could be a problem, because when the client try to get the cluster's topology, it will get the brokers ' private addresses , which is not reachable.
Regards,
Kien
On Aug 11, 2017, at 18:18, "Tzu-Li (Gordon) Tai" <[hidden email]> wrote:
|
It is solvable? I'm not an expert of this stuff and the cluster is managed by the lab responsible. Maybe I can ask him to do something in order to solve.
|
Hi, I don’t have experience running Kafka clusters behind proxies, but it seems like the configurations “advertised.host.name” and “advertised.port” for your Kafka brokers are what you’re looking for. For information on that please refer to the Kafka documentations. Cheers, Gordon On 12 August 2017 at 4:28:41 PM, AndreaKinn ([hidden email]) wrote:
|
Free forum by Nabble | Edit this page |