Re: [Postgres-xc-developers] Support for SET command

I understood.   Thank you for reminding.
----------
Koichi Suzuki



2011/4/4 Abbas Butt <abb...@te...>:
> In fact Michael is doing this in an effort to clean regression failures.
> Perhaps we can have a discussion on what work would be required to make a
> test case pass, and then prioritize them and work on them according to
> priority.
> Regards
> Abbas
> On Mon, Apr 4, 2011 at 6:29 AM, Koichi Suzuki <ko...@in...>
> wrote:
>>
>> I have just a concern on this series of "SET" command discussion.
>>
>> Last year, this was assigned to Andrei but after the review to prioritize
>> what to do next in March, we've dropped this from this fiscal years issue
>> list because others seemed to be more important.
>>
>> Should we give higher priority to this?   Then what should we trade?
>>
>> Regards;
>> ---
>> Koichi
>>
>> On Sun, 3 Apr 2011 05:33:27 +0900
>> Michael Paquier <mic...@gm...> wrote:
>>
>> > Hi all,
>> >
>> > I wanted to have everyone's opinion about what to do for the support of
>> > command SET.
>> > I am pretty preoccupied about this support as it has a lot of
>> > consequences
>> > on regression tests.
>> >
>> > Here are the set commands:
>> > SET ROLE, SET CONSTRAINT, SET
>> >
>> > There are also options to do a set for LOCAL (set by the end of current
>> > transaction) and SESSION (set as long as the session as alive).
>> > As far as I could see, it is necessary to make the management of SET
>> > through
>> > the pooler.
>> > When a SET command is issued, the pooler has to pick up the command and
>> > apply it to each connection of the pooler agent holding connections to
>> > backends.
>> >
>> > For local parameters, such parameters have to be changed when
>> > transaction is
>> > committed, and now XC does not do anything with pooler at commit... So
>> > perhaps it may be OK not to do anything for local parameters for the
>> > time
>> > being but just to focus on session parameters. This is at least OK for
>> > the
>> > first implementation step I think.
>> >
>> > Question: is there a libpq API or something that may help to pass down
>> > the
>> > parameter through the connection to the backend?
>> > And with this API the session parameter is alive as long as the
>> > connection
>> > is not cut.
>> > Pooler holds a connection pointer. I was thinking about putting a kind
>> > of
>> > flag on the pooler agent that would permit to identify which agent had
>> > been
>> > modified by a SET command.
>> > With this flag, it may be easy to manage session parameter. Isn't it
>> > enough
>> > to cut the connections on pooler that have been modified by session
>> > parameters when session is cut instead of putting them back to pooler?
>> >
>> > The problem is what can be used to modify a session parameter for a
>> > specific
>> > connection that is hold by the pointer?
>> > Does anyone have any idea or suggestion?
>> > --
>> > Thanks,
>> >
>> > Michael Paquier
>> > http://michael.otacoo.com
>>
>>
>> ------------------------------------------------------------------------------
>> Create and publish websites with WebMatrix
>> Use the most popular FREE web apps or write code yourself;
>> WebMatrix provides all the features you need to develop and
>> publish your website. http://p.sf.net/sfu/ms-webmatrix-sf
>> _______________________________________________
>> Postgres-xc-developers mailing list
>> Pos...@li...
>> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>
>
> ------------------------------------------------------------------------------
> Create and publish websites with WebMatrix
> Use the most popular FREE web apps or write code yourself;
> WebMatrix provides all the features you need to develop and
> publish your website. http://p.sf.net/sfu/ms-webmatrix-sf
>
> _______________________________________________
> Postgres-xc-developers mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>
>