If you have a system with a fixed number of zones, you can distribute the total amount of memory to the zones except what you want to reserve. If you don't know how many zones you will have or on which system your zone will be run, you can't easily manage how many memory you will cap to your zone. Also, when you use memory cap, you have to do it for each zone and it's a pretty workload when you have 800 zones in production. It could be better to have, let say, a kernel parameter to reserve memory. It's a reverse thinking. The zones won't be allowed to go up to the memory cap but not beyond what it will be reserved for the global zone.

Thanks for the discussion we have,

Pascal


Jerry Jelinek a écrit :
Pascal Fortin wrote:
Hi Jerry,

Thanks for your response. This is not overprovisioned system but you can have a bug somewhere that lead to use all memory and, of course, you have to correct it. But all other zones in the system will suffer from this situation and in consolidation context, it's annoying. As the zones are moving from one system to another, it's not so easy to use memory capping. This is why a reserved memory could be more intelligent and practical.

I'd like to understand what issues you've seen
with memory caps and zone migration.  Can you
explain your problem further and explain how
memory sets would solve it?

Thanks,
Jerry

--


Pascal FORTIN
Services Account Manager

Sun Microsystems France
13 avenue Morane Saulnier
78140 Velizy Villacoublay

Phone x30401 / +33 1 34 03 04 01
Mobile +33 6 85 83 10 01
Email pascal.for...@sun.com



_______________________________________________
zones-discuss mailing list
zones-discuss@opensolaris.org

Reply via email to