- Original Message -
From: "Jeroen T. Vermeulen" <[EMAIL PROTECTED]>
> I ran into the same need (Bruce, we discussed this at FOSDEM in Brussels
> this February) for libpqxx. My code tries to compensate for the
> possibility that the backend connection is lost while waiting for a reply
>
"Christian Plattner" <[EMAIL PROTECTED]> writes:
> From: "Tom Lane" <[EMAIL PROTECTED]>
>> How much later? clog is not kept forever.
> As far as I understand your code there is only one chance that information
> in clog gets lost:
> If XIDs are reused then ExtendCLOG will overwrite existing entri
"Christian Plattner" <[EMAIL PROTECTED]> writes:
> From: "Tom Lane" <[EMAIL PROTECTED]>
>> How much later? clog is not kept forever.
> ... Ofcourse one should not do a
> VACUUM FULL while not being sure about the status of a transaction in the
> past :)
As long as you haven't done a cluster-wide
On Fri, Jun 20, 2003 at 02:41:29PM +0200, Christian Plattner wrote:
>
> Well, my client (actually it is a middleware layer which routes transactions
> to a set of replicas) keeps its own log, because it must be able to handle
> arbitary failures. So it never looses its state.
In that case perhaps
- Original Message -
From: "Jeroen T. Vermeulen" <[EMAIL PROTECTED]>
> > btw: There could be also other reasons for the client to loose the
> > connection (i.e. client process crashes).
>
> In that case the client would lose all its state as well, so not really
> a problem that can be ha
- Original Message -
From: "Tom Lane" <[EMAIL PROTECTED]>
> How much later? clog is not kept forever.
I took a deeper look into the source. Forget my last posting.
As far as I understand your code there is only one chance that information
in clog gets lost:
If XIDs are reused then Exte
On Fri, Jun 20, 2003 at 10:20:14AM +0200, Christian Plattner wrote:
>
> Well, I wouldn't rely solely on TCP when assuring consistency. Also, I don't
> think that the backend will ever inspect its TCP socket while committing.
No, but its underlying IP stack would.
> btw: There could be also oth
- Original Message -
From: "Jeroen T. Vermeulen" <[EMAIL PROTECTED]>
> > I see a race condition in this approach: if you reconnect too fast, and
the
> > backend which actually should commit is still in progress (assume it
takes a
> > while to commit for whatever reasons) you get the impr
On Fri, Jun 20, 2003 at 09:35:08AM +0200, Christian Plattner wrote:
>
> I see a race condition in this approach: if you reconnect too fast, and the
> backend which actually should commit is still in progress (assume it takes a
> while to commit for whatever reasons) you get the impression that it
- Original Message -
From: "Tom Lane" <[EMAIL PROTECTED]>
> How much later? clog is not kept forever.
Due to my setup, I could assure, that for the XID I ask for always
(ShmemVariableCache->nextXid - XID) < C (and C is in my case something
around 150).
holds. A possible solution co
- Original Message -
From: "Jeroen T. Vermeulen" <[EMAIL PROTECTED]>
> I ran into the same need (Bruce, we discussed this at FOSDEM in Brussels
> this February) for libpqxx. My code tries to compensate for the
> possibility that the backend connection is lost while waiting for a reply
>
Hi all,
I am currently implementing an experimental middleware based replicator for
a set
of fully replicated databases.
Do be able to handle all sorts of failures I needed two functions:
- A function to get the current XID
- A function which I can use later to tell if a given XID
commited/abort
On Thu, Jun 19, 2003 at 05:16:10PM +0200, Christian Plattner wrote:
>
> Do be able to handle all sorts of failures I needed two functions:
>
> - A function to get the current XID
> - A function which I can use later to tell if a given XID
> commited/aborted/whatever
I ran into the same need (Bru
"Christian Plattner" <[EMAIL PROTECTED]> writes:
> Do be able to handle all sorts of failures I needed two functions:
> - A function to get the current XID
> - A function which I can use later to tell if a given XID
> commited/aborted/whatever
How much later? clog is not kept forever.
Hi all,
I am currently implementing an experimental middleware based replicator for
a set
of fully replicated databases.
Do be able to handle all sorts of failures I needed two functions:
- A function to get the current XID
- A function which I can use later to tell if a given XID
commited/abort
15 matches
Mail list logo
