*@Fabian do you register any types / serializers via
ExecutionConfig.registerKryoType(...) /
ExecutionConfig.registerTypeWithKryoSerializer(...)?*
Nope, not at all. our flink job code has nowhere the word "Kryo" at all.
thx for looking into it ...
--
*Fabian WollertZalando SE*
E-Mail: fab...@
I quickly checked the implementation of duplicate() for both the
KryoSerializer and StreamElementSerializer (which are the only serializers
involved here).
They seem to be correct; especially for the KryoSerializer, since
FLINK-8836 [1] we now always perform a deep copy of the KryoSerializer when
d
No, not yet. We lack some knowledge in understanding this. The only thing
we found out that it happens most probably in the Elasticsearch Sink,
because:
- some error messages have the sink in their stack trace.
- when bumping the ES nodes specs on AWS, the error happens less often (we
haven't bumpe
Any news on this? Have you found the cause of the error?
On Fri, Jun 28, 2019 at 10:10 AM Flavio Pompermaier
wrote:
> Indeed looking at StreamElementSerializer the duplicate() method could be
> bugged:
>
> @Override
> public StreamElementSerializer duplicate() {
> TypeSerializer copy = typ
Indeed looking at StreamElementSerializer the duplicate() method could be
bugged:
@Override
public StreamElementSerializer duplicate() {
TypeSerializer copy = typeSerializer.duplicate();
return (copy == typeSerializer) ? this : new
StreamElementSerializer(copy);
}
Is ti safe to return
Hi Fabian,
we had similar errors with Flink 1.3 [1][2] and the error was caused by the
fact that a serialised was sharing the same object with multiple threads.
The error was not deterministic and was changing from time to time.
So maybe it could be something similar (IMHO).
[1] http://codeha.us/a
additionally we have these coming with this as well all the time:
com.esotericsoftware.kryo.KryoException:
java.lang.ArrayIndexOutOfBoundsException
Serialization trace:
_children (com.fasterxml.jackson.databind.node.ObjectNode)
at
com.esotericsoftware.kryo.serializers.ObjectField.read(Obj
Hi, we have some Flink Jobs (Flink Version 1.7.1) consuming from a Custom
Source and Ingesting into an Elasticsearch Cluster (V.5.6). In recent
times, we see more and more Exceptions happening like this:
com.esotericsoftware.kryo.KryoException: Unable to find class: com. ^
at
com.esoteric
