It's an external XA Manager (Wildfly) that coordinates transactions between a
message queue and the Oracle database.
Thomas
Mark W. Farnham schrieb am 07.10.2020 um 23:47:
Two phase XA transactions between what and what?--
Oracle 2 phase commit amongst multiple Oracle data stores is one of the
sweetest bits of code ever written (thanks, Gary Hallmark, and XA is most
kindly described as a camel built by committee when you wanted a race horse.)
What are they trying to accomplish that involves both hibernate and XA, and
how do you decontaminate your brain after the experience (assuming wildly
that it ever comes to completion)?
Very often not even the 2 phase commit on Oracle is required, but it could be
if you're trying to lockstep update something like inventory quantities or
bank balances on two separate primary databases which can both initiate
changes. If it is just a recording feed, that is probably NOT a 2 phase
commit requirement even if the system sending the update requires commit
confirmation to update a status locally.
Good luck.
mwf
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Thomas Kellerer
Sent: Tuesday, October 06, 2020 7:22 AM
To: oracle-l@xxxxxxxxxxxxx
Subject: Re: Excessive "latch: row cache objects" waits with simple INSERT
statement
Yes, looks like it. Apparently Hibernate lies when it claims it's doing
batched insert.
No VPD, no triggers.
But I just learned that those INSERTs are done using a 2-phase commit (XA
Transaction).
Could that be the cause these latch waits?