Ok,
That may have been a bit too much :-) However, it was useful.
There seem to have several possible avenues:
1) You are using SolrJ and your SolrJ version is not the same as the
version of the Solr server. There was a bunch of things that could
trigger, especially in combination with Unicode but also with - at
some point - SolrJ sending javabin format to an XML endpoint. So, I
would check that first. I know you said that it works with nonCluster
setup, but are you definitely using the same client/configuration in
both approaches? Especially because of the following line below:
DEBUG - 2018-10-19 02:15:23.384;
org.apache.solr.update.processor.LogUpdateProcessor; PRE_UPDATE FINISH
{{params(update.contentType=3Dapplication/xml),defaults(wt=3Djavabin&version=3D2)}}
2) If that did not help, I would focus on whether it is any document
to Thai core that causes an issue or just specific one. If you are
doing the tests, make sure to commit each time, because of (3 below):
3) It is possible that there is a unicode related bug in the
inter-cluster communication. Possible. But I could not find the
specific Jira. Still, if you cannot reproduce it on later version of
Solr, that is a likely scenario.
4) If all else fails, use something like Wireshark and capture the
network-level traffic during this error. This will show you exactly
what is being passed around (https://www.wireshark.org/). This is
using a power-hammer on a nail, but - if you read this far - I suspect
you are out of other options.
Let us know when you resolve the issue, for those who may see it later.
Regards,
Alex.
On Sun, 21 Oct 2018 at 12:16, Moshe Recanati | KMS <mos...@kmslh.com> wrote:
>
> Hi,
>
> Thank you.
>
> Full stacktrace below
....
