Client dies in transaction ?
How can I address the problem of table locking/rollback when my client app
dies in the middle of a transaction process?
TIA
Roberto
--
Roberto Moreda
Resp. Dpto. Inform�tica Handem/San Luis
Tlf +34 981 779000
Fax +34 981 779022
Pol. Piadela Sur, Autov�a A6 Sal.567
15300 Betanzos (A Coru�a) - Espa�a
Roberto Moreda wrote:
How can I address the problem of table locking/rollback when my client app
dies in the middle of a transaction process?
That question has been asked by me :) By inhumane and barbaric
experiments I found that backend successfully rolls back
transactions from dead clients. The only problem is that if
there was a process of creating new tables (or other stuff
like that), there remains a corpse of that semi-created table.
But someone said this problem should be addressed, or am I
mistaken? :)
Roberto has already reproduced two of three of my Big Questions:
speed of joins and sudden client death. :) The third question
should be: what do I do with transaction deadlock, since there
is no lock wait timeouts?
--
Leon.
---------
"This may seem a bit weird, but that's okay, because it is weird." -
Perl manpage.
Import Notes
Resolved by subject fallback
Leon wrote:
speed of joins and sudden client death. :) The third question
should be: what do I do with transaction deadlock, since there
is no lock wait timeouts?
Server should abort one of transaction to resolve deadlock -
we don't use timeouts for this.
Did you get unresolved deadlocks?
If so then there is a bug in lock manager.
Vadim