There were some race conditions where one transaction releasing a lock could
miss the wakeup of another transaction waiting for the lock. The result is
that the waiting transaction would be stuck until tokudb_lock_timeout
expires.
Maybe that pull request could help in this case also.
The stack trace shows TokuDB waiting for a row lock.
There is a pull request for fixing some bugs with similar symptoms as this:
https:/ /github. com/percona/ PerconaFT/ pull/360
There were some race conditions where one transaction releasing a lock could
miss the wakeup of another transaction waiting for the lock. The result is
that the waiting transaction would be stuck until tokudb_lock_timeout
expires.
Maybe that pull request could help in this case also.