Re: IGNITE-2779 BinaryMarshaller caches must be cleaned during client reconnect.

[Nikita Amelchev <ns...@gmail.com>]

Hi, Igniters.

I have done issue 2779 - BinaryMarshaller caches must be cleaned during
client reconnect.[1]

Could you, please, review?

Solution:

We stop server nodes and keep alive the client. Then we take a new instance
of the cache and client node throws an exception in
BinaryReaderExImpl.getOrCreateSchema when try read from cache. A client
couldn't take metadata from context.

Problem was that user type descriptor was cached on the client and absent
on servers. I clean user descriptors after client disconnect. It fixes the
issue. I have tested solution.[2]

[1]: https://issues.apache.org/jira/browse/IGNITE-2779

TC:
https://ci.ignite.apache.org/project.html?projectId=Ignite20Tests&tab=projectOverview&branch_Ignite20Tests=pull%2F2518%2Fhead
Upsource: https://reviews.ignite.apache.org/ignite/review/IGNT-CR-328

2017-02-16 17:22 GMT+03:00 ALEKSEY KUZNETSOV <al...@gmail.com>:

> will take the ticket https://issues.apache.org/jira/browse/IGNITE-2779
> --
>
> *Best Regards,*
>
> *Kuznetsov Aleksey*
>



-- 
Best wishes,
Amelchev Nikita