Comments on: Statspack on RAC

By: Timur Akhmadeev

Timur Akhmadeev — Mon, 17 Jan 2011 08:14:56 +0000

From programming point of view the lock approach is much more clear and does exactly the thing it supposed to do. Chains, on the other hand, introduce ordering which seems to be unnecessary in this situation. Code to set up a chain IMO isn't complicated though.

I’m exaggerating for effect, of course, but not by much

Well, indeed this code (like any other non-RAC friendly piece of software) will put some pressure on private interconnect and LMS processes. So, in theory, it could compete with applications running in the cluster. But I think there're chances such situations could pass without severe affect on other applications. So "lock up the system" is indeed an extreme scenario - and I wouldn't mind knowing other details behind this sentence (with numbers). Also thank you for WF enqueue description. I've seen it multiple times (with quite huge wait time reported) but didn't realize what Oracle tries to sync.

By: Jonathan Lewis

Jonathan Lewis — Sun, 16 Jan 2011 17:45:47 +0000

Tim, Thanks for that thought. It look's like I fell into the old trap of "repeat whatever worked last time" :( If you've got an example of scheduler job chain setup on your website feel free to add a link to it here.

By: Tim Hall

Tim Hall — Fri, 14 Jan 2011 20:26:45 +0000

Hi.

If you are using 10.2 or above you could also do this using a scheduler job chain. Having said that, your sample code is pretty concise and the scheduler job chain setup probably looks a little more complicated if you are not used to using it. :)

Cheers

Tim…