Re: Data corruption problem...

I haven't seen any replies to this, so I will,
although I am not a j2ee/jdbc person.

On fim, 2006-11-30 at 14:14 -0700, Nathan Wilhelmi wrote:

> Server: 8.0.3 on Solaris 9
> JDBC Driver: 8.0.311
...
> delete from X where id in (select bad_id from Z where name='qwerty');
> delete from Y where id in (select bad_id from Z where name='qwerty');
> delete from Z where name='qwerty');
...
> Now we have J2EE (Tomcat hosted) app that access this database via
> connection pool / JDBC driver.

tomcat version ?

> 1) Start the J2EE app and do some work to access the database, although
> I don't think this matters.
> 2) Shut down the app server, and as result the connection pool starts to
> shut down.
> 3) Once the connection pool has started to shut down execute the
> statements listed above from PGAdmin.
> 4) PGAmin will hang for a second like it's waiting for locks.
> 5) The statement will return normally, reporting the rows affected from
> the last statement.
> 6) Tables X and Y are now empty. This is not good.
> ...
> Has anyone seen this before, any suggestions how to debug
> this or where to start looking?

I would start with enabling statement logging, and see
exactly what SQL the server is executing during the
scenario.

gnari