Hello Sebastian, <property name="openjpa.jdbc.DBDictionary" value="batchLimit=100,tableType=myisam"/> was added to mysql_persistence.xml to speed up things. While Vasily did migration from Hibernate to JPA he report the performance degradation. I did investigate this and found this is because of InnoDB used by default.
Initial language import was significally slower. My current results are MyISAM: initial lang import == 9 sec; subsequent import == 0/1 sec InnoDB: initial lang import == 3 sec; subsequent import == 0/1 sec was measured with ./admin.sh -l Surprisingly right now InnoDB was faster, can you recheck these results on your machines? I would vote for hard delete of everything or add option to restore/purge deleted things. Currently it will be hard to implement cascade delete since our object structure is not good for it (all object need to have references to connected entities, we have "Long ref_id" instead :( ) On Sat, Sep 22, 2012 at 3:24 PM, [email protected] < [email protected]> wrote: > Hi Maxim, > > I saw your last commit, still studying it :) > > One thing we might should consider is using InnoDB as default, as it > supports foreign key constraints. > For example the delete Organisation cannot work, there are references > in the table "rooms_organisation". > While we are currently using MyISAM we do not recognizes it, if > switching to InnoDB or Postgres you migth immediately see that it does > not work. > > We might also discuss which of the tables we are going to actually > really delete and which ones we only flag as deleted. > I think organizations can be deleted "Hard". > Further "Hard delete": configurations, ldapconfigs, servers > But it will not work for users, rooms. > It might makes sense to argue that rooms should be deletable to free > up server space cause a deleted room has assigned room files that you > might want to get rid of to free your disk. However there are also so > many other referenced tables with records that should not be deleted > at all when deleting a room record so that I doubt that it makes sense > to put time into this complexity now. > > What do you think? > > Sebastian > -- > Sebastian Wagner > https://twitter.com/#!/dead_lock > http://www.webbase-design.de > http://www.wagner-sebastian.com > [email protected] > -- WBR Maxim aka solomax
