|
(I've used LVS some 6 years ago, now I'm back :-D)
I'm looking at this page...
http://www.linuxvirtualserver.org/docs/scheduling.html
...and I think the algorithms are pretty well explained. But what I need
is to customize the job scheduling based on my own criteria.
E.g., I've a bunch of servers doing all of them pretty much the same
thing and I want to distribute the load based on criteria such as:
- disk usage
- OS load average
- CPU usage
- custom hooks into my own software
- You Name It (TM)
This would require LVS to query the servers somehow (SMTP, ssh, or
something like that).
Then, if I understand LVS correctly, a script should tweak the relative
weights of each server.
Is that correct?
Am I missing something?
Any preferred algorithms for such a scenario?
Any issues with tweaking the weights all the time?
Any guidelines, best practices, common pitfalls, etc., when doing such a
thing?
Thank you,
--
Florin Andrei
http://florin.myip.org/
|
