Comment 3 for bug 1734312

commit 12a9e13401b4d9e7b781034eddf04ffe957c07d0
Author: Krunal Bauskar <email address hidden>
Date: Wed Nov 29 17:17:22 2017 +0530

    - PXC#806: Early read of query-id result in stale trx handle assignment.

      * During an execution of a Stored Procedure (SP) as a background
        action (say through event scheduler), query-id get assigned at
        later stage, before PXC captures it to derive wsrep_next_trx_id.

      * PXC uses this wsrep_next_trx_id as unique id to maintain
        TrxHandle and thread mapping.

      * This early capture of query-id resulted in assigning an invalid
        wsrep_next_trx_id (read 0). Problem starts if multiple threads
        gets the same invalid (read 0) id as their unique id to look
        for TrxHandle that eventually cause sharing of TrxHandle.

      Solution:
      --------

      * Fixed by ensuring that when PXC reads query_id, it is a valid id.
        If not, then a valid query_id is generated.
        (MySQL flow anyway generates it but at later stage).

      * We could have also corrected it by skipping setting of
        wsrep_next_trx_id at the said stage and waiting for MySQL to
        generate it but that means a drill-down code inside
        the main logic. So keeping it at outer layer
        makes it more manageable.