Here's a thought I had...
With large tables the default value of MaxNoOfConcurrentOperations
isnt' sufficient. This means that ALTER TABLE FOO ENGINE=NDBCLUSTER
will always fail.
Same issue with TransactionDeadlockDetectionTimeout ...
We weren't able to ALTER our tables without increasing
TransactionDeadlockDetectionTimeout...
That said maybe the correct upgrade procedure would be to either
increase these values temporarily OR do a mysqldump.
The problem is that mysqldump creates one large INSERT which is
problematic as well. If it could split up the insert into 100 row
batches that would be ideal.
Thoughts? Is there any ideal solution to this problem?
Kevin