| Re: [eclipselink-users] EclipseLink data-source usage |
On 10/7/2014 10:33 AM, Rick Curtis wrote:
On the contrary, the object read through read connection lands in L2 cache, therefore next time find is called (unless refresh query hint is specified) it fetches the object directly from the cache (= no trip to the db).I don't think this matters, but I was running with <shared-cache-mode>NONE</shared-cache-mode>.
To summarize what you said... until something is written to the transaction the read connection will be used? This seems very inefficient because in my case, making 5 subsequent trips to the database will result in 5 connections being opened/closed. Connection pooling helps with some of this performance hit, but pools are still slower than no pool.
If you don't want to use shared cache (or just always want to reads through the write connection) there is a couple of alternative options:
1. specify puProperty "eclipselink.jdbc.exclusive-connection.mode" "Always";
2. specify puProperty "eclipselink.transaction.join-existing" "true" (note that also could be done on per entity manager basis)
Can I just not configure a non-jta-data-source or are there any scenarios that require use of a non-jta-data-source?
Thanks,Rick
On Tue, Oct 7, 2014 at 9:22 AM, andrei ilitchev <andrei.ilitchev@xxxxxxxxxx> wrote:
The described scenario is default one (exclusive connection is not).
When nothing has been written yet inside transaction find by pk is done in L2 cache first, if not found there - in the db, but through reading connection (and result merged into L2 cache).
On 10/7/2014 10:13 AM, Rick Curtis wrote:
I have been doing some testing on EclipseLink around datasource usage (jta vs non-jta) and there is a scenario that I don't quite understand.
It scenario involves a SLSB method that has @TransactionAttribute(TransactionAttributeType.REQUIRES_NEW). I found that if inside of that method I do a number of em.find(..) operations, each operation gets a new (non-jta) connection. If I start the method with an em.persist(...);em.flush(); followed by a number of em.find(...) operations all of follow on em.find(...) calls use the one connection(jta). From what the documentation states (eclipselink.jdbc.exclusive-connection.mode), the default EclipseLink behavior is to send all reads/writes through an exclusive connection when running inside of a transaction. What am I missing?
Are there any best practices around the use of jta/non-jta datasources when using a JTA EntityManager?
Thanks,Rick--
Rick Curtis
_______________________________________________ eclipselink-users mailing list eclipselink-users@xxxxxxxxxxx To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/eclipselink-users
_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/eclipselink-users
--
Rick Curtis
_______________________________________________ eclipselink-users mailing list eclipselink-users@xxxxxxxxxxx To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/eclipselink-users
_______________________________________________
eclipselink-users mailing list
eclipselink-users@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/eclipselink-users