Hi John,
To answer your question the short way, I can see 2 solutions : a) Modify the
application so that the database itself is cut into smaller databases. That
would implement smaller control blocks, with more of them to queue on, thus
reducing the contention on a single resource, which it seems is what your
users actually experience. Or b) significantly upgrade your system, so that
the transactions queuing on a given control block are processed faster. Or
do both. The problem is, each solution has its price. Option a) also
probably means some development time. Depending on where your application is
in its lifecycle, you may or may not want to change anything.
Now, let's think a little :
- Have you had this problem for a long time ? If yes, how did you cope with
it until yesterday ?
- If the problem is new, did you change anything ? A new application
somewhere ? New users on an existing application ? Some change in the
database schema ? Well, anything you changed on that system may be relevant
- Can you tell if impeded users are actually queuing on the SAME database
control block ? Or different databases ?
- Do users have a tendency to logoff all at the same time, e.g. end of shift
? Did they always have the same behaviour, or was there some recent change ?
- Can you tell whether your system experiences memory pressure ? Lots of
swapping IOs ?
I hope these few ideas can help. Don't hesitate to get in touch with me.
Best regards,
Christian "still trying to help, after all those years" Lheureux
> -----Message d'origine-----
> De : HP-3000 Systems Discussion [mailto:[log in to unmask]] De la part
> de John Clogg
> Envoyé : lundi 14 janvier 2008 23:53
> À : [log in to unmask]
> Objet : [HP3000-L] DBCLOSE Causing Impedes
>
> We have a problem each day when a group of users log off the MPE system at
> the end of their shift. As these employees shut down their sessions, we
> start seeing lots of sessions spending a significant amount of time in an
> impeded state. It affects the sessions that are logging off, as well as
> several others. When I have displayed stack traces of impeded sessions in
> Glance, it seems they are usually doing a DBCLOSE. I surmise that there
> is contention for a database control block, which causes these processes
> to impede one another. Can anyone suggest a solution to this problem? Is
> there any kind of setting that would allow greater concurrency in the
> DBCLOSE process or reduce the amount of time each process has the resource
> locked? Does anyone disagree with my diagnosis of the issue? Any
> suggestions would be appreciated.
>
> Regards,
> John Clogg
> _________________________________________________________________
> Get the power of Windows + Web with the new Windows Live.
> http://www.windowslive.com?ocid=TXT_TAGHM_Wave2_powerofwindows_012008
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
