Re: SQL-08S01 error in FreeTDS/Azure when executing stored proc for a few minutes from Linux (but not from Windows)

2012/11/23 James K. Lowden <[email protected]>:
> On Wed, 21 Nov 2012 22:53:44 +0000
> Frediano Ziglio <[email protected]> wrote:
>
>> Yes,
>>   however MS requirements are more strictly that Unix. By default many
>> systems starts sending keepalive after 2 hours idle, much more that 1
>> minute!
>
> Do you have a reference for Microsoft's keepalive requirements?
>
> I remember we had a problem with a long-running idle watchdog process
> that would occasionally end with "read from SQL Server failed".  I
> never determined why.
>

Yes, have a look at
http://blogs.msdn.com/b/sql_protocols/archive/2006/03/09/546852.aspx

However this is confirmed by the settings. They should be settings on
server side so client "could" avoid them but in case server was not
configured in such way the MS load balanced would hit and do its job.

The configuration at end seems
- idle timeout (first keepalive after idle) 30 seconds
- interval (time it try to send another keepalive if acknowledge is
not received) 1 second
- keepalive count (number of retry without acknowledge before disconnection) 5.

The default under Linux is 7200, 75, 9 while on FreeBSD is 7200, 75, 8.

Frediano