|
| Re: Time Required To Abort Transaction After Tuxedo Transaction Time |
 |
Thu, 24 Apr 2008 08:57:08 -070 |
Hi Ajeet,
I'm still not entirely clear what you want the client to see or to have to do.
If the client is starting the transaction with a tpbegin(30), then the client
should receive a TPETIME error 30 seconds (plus up to SCANUNIT additional
seconds as timeouts are only checked every SCANUNIT).
On the other hand, if the tpbegin is in the service implementation, it will
first notice the TPETIME when the next service invocation it makes complete, or
whenever it calls another messaging routine. The client won't get notified of
the transaction timeout until such time as the service implementation returns,
or until the blocking time (again plus up to SCANUNIT additional seconds) is
exceed. Service blocking time can be controlled on a per call, per context, per
service, or per domain basis with Tuxedo since version 9.1.
Once the blocking timeout has occurred, the client will receive a failure
notification. Depending upon the service implementation though, it's possible
that the client will receive a service timeout from blocking timeout and the
server still complete the request successfully. How much of an impact that is
to your application is something you'd need to determine.
I hope this helps.
--
Regards,
Todd Little
|
| Post Reply
|
|
|
|
|
|
|
|
|
|