Yes, there are some issues where we share connection handles in 1.x
database drivers. In the ccscript3 module I tried playing with where
connections occur and how sessions are managed to reduce or eliminate
this kind of blocking and maybe some of that should be backported. But
having a seperate connection from each trunk is also workable. My only
concern is with overhead on high port density systems, though it
certainly is better than the situation it is replacing :).
Etoile Dièse wrote:
Hello,
In the version 1 of Bayonne, when a SQL request is long, or blocked for
example because the table is locked by another request, if the user hangs up,
no more mysql request can be done and every trunk doing it remains blocked.
The reason is that the mysql module object is unique and protected by an
internal mutex. The mutex is positionned before the operations and when the
request is suddenly interrupted by the hangup, the mutex stays in a blocked
state.
We modified the file mysql.cpp for using a private mysql connection handler
for every trunk that needs it. I will post the patch against the 1.2.14
version as soon as we have tested it (this week-end).
begin:vcard
fn:David Sugar
n:Sugar;David
org:GNU Telephony
adr:;;;;;;USA
email;internet:[EMAIL PROTECTED]
x-mozilla-html:FALSE
url:http://www.gnutelephony.org
version:2.1
end:vcard
_______________________________________________
Bayonne-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/bayonne-devel
