Re: ndb cluster 7.4.12 seems to limit the number of tables/database to 50 when character set is utf8mb4

[Mikael Ronström]

I went through all the configuration variables and found
one potential configuration parameter that might affect this.

It is StringMemory. This is a percentage parameter. So in principle
it decides how much memory is allocated for attribute names, table names
and defaults on attributes. It allocates a fair amount of memory for this
purpose based on the setting of the above mentioned configuration
parameters.

Next it multiplies the size by a percentage given in StrongMemory.
By default this is 25. So setting this to a higher value provides
more StringMemory.

So most likely using utf8mb4 meant that you needed much more memory
for attribute names, table names and default values and you
probably ran out of the StringMemory resource. I presume it should be
possible to get some kind of warning printout that defines more clearly
what resource was missing.

The error code for this error is OutOfStringBuffer and has number 773.

Why the database runs slower with utf8mb4 I am no expert in, but there
are more characters to process and probably more complex conversions
in various places in the code.