Re: Deadlock condition in driver

Previously we did not want to move to nio to be able to support older JDK
versions, but with the next major release we're ready to drop that
support. If you have this code, I'd certainly be interested in looking at
it. One of the big drawbacks of breaking up a single execute call into
multiple batches is that in autocommit mode you would get one transaction
per batch instead of the previous behavior of one transaction for the
whole execute call.

Kris Jurka

On Thu, 22 May 2008, Richard Evans wrote:

> We hit this problem and modified the driver to use non-blocking IO to fix it.
> You lose some of the more unusual features of the driver (such as switching
> the socket after open), but it does fix the blocking.
>
> Richard Evans
>>
>>
>> On Wed, 21 May 2008, Daniel Migowski wrote:
>>
>>> I came across a deadlock condition in the JDBC driver that rises when very
>>> large queries, containing thousends of statements are send to the server
>>> with statement.execute().
>>
>> We already consider this case for batch execution and break the batch into
>> an internal size that we expect is safe from deadlock. It looks like we
>> should be doing the same for these one query batches. I'm not sure how
>> tough that will be, but I'll take a look.
>>
>> Kris Jurka
>>
>
> --
> Sent via pgsql-bugs mailing list (pgsql-bugs(at)postgresql(dot)org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-bugs
>