MySQLInterfaceError: Lock wait timeout exceeded; try restarting transaction

[Pete Ditmars]

Hi,

We are using MySQL NDB cluster version
mysql-cluster-commercial-server-8.0.25-1.1.el7.x86_64

We are using python clients with SQLAlchemy and Java clients with eclipselink as ORM layers.

Both kinds of clients are randomly encountering "Lock wait timeout" errors on insert, update, delete, in various tables. As far as we can tell there is zero contention for the tables. We have about 15 databases in a single mysql instance and it doesn't seem to matter which database or which table.

The documentation suggests the relevant configuration parameter TransactionDeadlockDetectionTimeout defaults to 1200ms as the "lock wait timeout", and is also used also for other cluster functions. We are not sure increasing this parameter is a valid solution. Neither, it would seem, is wrapping every attempt to commit a transaction with a retry.

These lock wait timeouts seem to be increasing in frequency in our environment.

How can I troubleshoot this further to find out why we are encountering this?

Is it possible there is some inter-node communication issue that is causing these?

Thanks,

Pete Ditmars