Hi Ricardo,
To use of Sequoia + Postgres + sequence you may have the follow problem
that I discovery today:
When you make a insert that use sequence in Postgres, and the insert
fail, the sequence increment will never rollback.
Than I discovery a problem with this scenario:
- I have a backend_1 and backend_2;
- Make a backup of backend_2 (the backend go to disabled state);
- Make a sequencial insert that will fail while backend_2 during the
backup (the insert will execute only in backend_1, the sequence in the
backend_1 will be incremented but the insert will not occurs);
- The backup over, and backend_2 become enabled (the recovery log will
replay the insertions and update that occurs during the backup, but will
NOT replay the insertion that failed and increment the sequence in the
backend_1);
- So the sequence in the 2 backends are inconsistent, and in the next
insertion the data will be different in the backends;
To workaround this problem you may use stored procedures to insert all
the datas that use sequencial values, because even if the procedure fail
it will be replayed by the recovery log.
I'm using the Sequoia, and don't know if this happens in the 2.x.
Leandro Borges
Ricardo Kustner escreveu:
Hi Emmanuel,
Emmanuel Cecchet wrote:
Hi Ricardo,
(I wrote this to the community list earlier, but I guess that's the
wrong place?)
The community list is not read as widely as the Sequoia list. So for
Sequoia related questions, please use the Sequoia mailing list.
I'd suggest the mailinglist page gets updated with some more
information on the purpose of the community en sequoia list. The low
traffic on the community could make it look as if there isn't much
activity going on in the Sequoia community :)
We are using a Sequoia cluster with a hibernate based application in
tomcat.
One thing which confuses me, is what the best way is to have the
application communicate with Sequoia:
A) let hibernate use the tomcat jdbc resource.. or...
B) have hibernate connect directly to Sequoia through jdbc? The
reason I'm asking is that the sequoia docs talks about special tricks
to be used for using sequences properly
(http://sequoia.continuent.org/doc/2.10/userGuide/ar01s06.html, 6.6 and
6.7)
Would that work properly in situation (A) ?
The special tricks with sequences are related to PostgreSQL and they
apply to all cases if the application uses sequences. If you use
Hibernate to generate the database queries, it will use sequences
with PostgreSQL. Thus, you will need the trick mentioned in the doc.
If you are writing all SQL by hand and don't use PostgreSQL sequences
then you don't need the trick.
Thanks for your answer. I guess the developers of the application we
host will need to look into having their hibernate connect directly
with Sequoia and use the sequence trick.
Ricardo.
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
--
LEANDRO DE PAULA BORGES
Analista de Desenvolvimento de Software
PST Electronics
Phone: +55 19 3787 6402
[EMAIL PROTECTED]
www.pst.com.br
www.positron.com.br
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
