Probles with custom serializer and node disconnections

[Joan Pujol <jo...@joanpujol.cat>]

We have a cluster of several embedded nodes and from time to time we have
problems with "corrupted" caches. Usually when some of the nodes disconnect
for a long pause or similar reasons.

In the cache, we store domain entities using a custom serializer that
basically serializes de object to json and put the json content in a field
called "json".
Something as basic as:
write:

binaryWriter.writeByteArray("json", objectMapper.writeValueAsBytes(obj));

read:

objectMapper.readerForUpdating(obj).readValue(
binaryReader.readByteArray(JSON));


When we have the problem that cache it's "corrupted", the problem we have
is that when reading, we receive some objects that doesn't have the "json"
field.
And the schema it has is another that seems that is serialized using the
default serializer.
I've used the corrupted with quotes because the cache per see seems
correct, but some objects appear serialized with an incorrect schema.

Does someone have any clue of what can be happening? Is there any
possibility that when a node reconnects doesn't use our custom serializer
if it was configured to use it?

A lot of thanks in advance,

-- 
Joan Jesús Pujol Espinar
http://www.joanpujol.cat