FreeTDS Development Group

[Craig Berry <craigberry AT mac.com>]

On Tuesday, March 17, 2009, at 03:33PM, <Jeff.Goodwin AT fairchildsemi.com> 
wrote:
>freetds-bounces AT lists.ibiblio.org wrote on 03/17/2009 04:05:19 PM:

>The main changes are:
>
>OpenVMS VAX V6.2 - OpenVMS Itanium V8.3-1H1
>UCX V4.2 to HP TCP/IP Services V5.6 - ECO 3
>Digital Ada to GnatPro supplied by AdaCore
>Sybase client 10.0.1? to FreeTDS 0.82
>DECmessageQ V3.2 to BEA/Oracle MessageQ V5.0
>DEC C V5.6-003 to HP C V7.3-018 (just the wrapper program referenced 
>above, and of course FreeTDS)

Yikes.  It all should work in principle, but that's a lot of variables to 
have in play.  While FreeTDS is not the most likely source of your problem, 
it is one of the network-facing bits and can't be ruled out yet.  Most likely 
there's some subtle interaction of different pieces.

Here's a wild theory.  Do you have a keepalive setting that expires depending 
on whether anything shows up in your message queue within a particular time 
frame?  Say the database connection dies if it has been inactive for 4 hours. 
 As long as you have messages in the queue more frequently than that, you'll 
stay up, but if there is ever more than 4 hours between messages you'll get 
disconnected.  Do you have a way to control how often you receive a message 
in your test environment?

$ tcpip show protocol tcp

should tell you whether there have been any connections dropped by keepalive.

Can you reproduce the problem by simply logging in with tsql and letting it 
sit overnight or do you have to have your whole application/messaging setup 
running?

>We do have some Alphas, but it would be a major undertaking to try to get 
>this same environment running under Alpha as we do not have Alpha versions 
>of all the needed products.

Anything that could reduce the number of variables would be helpful.

>> Craig Berry, have you run freetds on Itanium?

Not yet, but I will soon.  We're finally replacing a very old Alpha sometime 
in the next few weeks.

>> I'm wondering if we might be hitting an Itanium-specific bug. 
>(Compiler?)

I think that's extremely unlikely, though of course not impossible.