ISPN-3937 Investigate the memory usage of JDBC cache stores when process...

[pruivo]

...ing a large result set

• Set the result type to TYPE_FORWARD_ONLY and the concurrency to CONCUR_READ_ONLY
• For the particular case of MySQL, setting the fetchSite to MIN_VALUE will fetch
  a single row at the time

https://issues.jboss.org/browse/ISPN-3937

[danberindei]

Shouldn't we force TableManipulation.getFetchSize() to return Integer.MIN_VALUE if the dialect is MySQL? We could try to log a warning if the user configured a custom fetch size, but I don't see a really good reason to want a custom fetch size on MySQL.

[pruivo]

@danberindei I didn't want to do that for the following reason: the db is there only for persistence and the dataset is small. IMO, this way allows the user to configure their need (it can fetch all table if hte dataset is small, or a single row for large dataset).

[danberindei]

@pruivo We wouldn't have any problem if MySQL really respected the fetch size :)

I would rather not write "If you're getting an OOM, please set fetchSize to -2147483648" in the user guide, the default value should be good enough in most cases.

Ideally, the default for MySQL should be {{Integer.MIN_VALUE}}, and if the user should be able to change it if he is sure his driver supports a custom fetch size and/or his dataset is small. But the table manipulation config doesn't know anything about the DB dialect, so implementing (and documenting) this would be pretty tricky.

[pruivo]

so, your point is to use MIN_VALUE when MySQL is used?

[danberindei]

Yes, didn't I say that? :)

[pruivo]

@danberindei sorry, to much multitask :)
PR updated!

[danberindei]

Thanks Pedro, integrated!