I'm observing a strange behavior using `KafkaSourc...
# troubleshooting
g
I'm observing a strange behavior using
KafkaSource
vs
FlinkKafkaConsumer
. My streaming application runs just fine using
FlinkKafkaConsumer
but when I use the
KafkaSource
API, it errors out saying, "unable to fetch topic metadata"! Anyone faced this issue before? The configuration to both clients are exactly the same.
m
Which version of Flink are you using?
g
1.14.3
m
I can only find Python related errors on that, are you using PyFlink?
g
No, its Java
m
Weird
The only thing that I can really find on that is that your Kafka service would be running low on memory; there's not an obvious thing what could be the case. We have done a number of improvements in the used Kafka Clients in later versions (1.15 and 1.16). I do think that upgrading would benefit you
g
Oh ok. Our Kafka cluster is super healthy - ~30% cluster CPU usage. Let me think on upgrading the clients. Will this upgrade have any impact on the cluster that will be running on
1.14.3
or do we need to upgrade cluster as well
1.15+
?
m
If you upgrade a job to a later version, you will also need to upgrade the cluster to the same version
g
Got it, will check it out. Thanks.