Re: BUG #5297: Add XATMI C API

Поиск
Список
Период
Сортировка
От Michael Musgrove
Тема Re: BUG #5297: Add XATMI C API
Дата
Msg-id 4DB697E1.1080506@redhat.com
обсуждение исходный текст
Ответ на Re: BUG #5297: Add XATMI C API  (Heikki Linnakangas <heikki.linnakangas@enterprisedb.com>)
Ответы Re: BUG #5297: Add XATMI C API  (Michael Musgrove <mmusgrov@redhat.com>)
Список pgsql-bugs
We would certainly be very interested in testing XA support. Our project
is called BlackTie - it's an XATMI implementation. We use the JBossTS
transaction manager for the transaction engine - although JBossTS is
written in Java, providing XA support was reasonably easy since it
conforms to the OTS specification which means we can control
transactions from a C based XA environment.

There is an example of using different resource managers within
transactions:
https://svn.jboss.org/repos/blacktie/trunk/atmibroker-xatmi/src/example/txfooapp
The files ora.c and db2.c show user code accessing resources from two
different vendors  - the XA specific information is defined in
btconfig.xml (basically it just exposes the XA open string, the symbol
name of the xa switch structure and the name of the library that exports
the switch symbol).

> On 21.04.2011 20:04, Michael Musgrove wrote:
>> I am referring to the X/Open specification Distributed Transaction
>> Processing: The XA Specification
>> (pubs.opengroup.org/onlinepubs/009680699/toc.pdf). If postgresql
>> provides a conforming Resource Manager switch then we can include
>> postgresql work along with interactions with other Resource Managers
>> (Oracle, DB2) in the same transaction. For example the equivalent
>> support in Oracle that I am asking for is documented at
>> http://download.oracle.com/docs/cd/B10501_01/appdev.920/a96590/adga1_xa.htm
>>
>
> When I wrote the patch for the backend support, I was not able to find
> any open source implementations of X/Open XA-compatible resource
> managers or transaction managers, which makes it kind of hard to
> implement the API correctly. There was plenty of JDBC XA
> implementations, however, so I used that for testing.
>
> If you can find something to test it with, it probably wouldn't be
> that hard to write the library.
>

В списке pgsql-bugs по дате отправления:

Предыдущее
От: Dave Page
Дата:
Сообщение: Re: BUG #5800: "corrupted" error messages (encoding problem ?)
Следующее
От: Michael Musgrove
Дата:
Сообщение: Re: BUG #5297: Add XATMI C API