On a system running SRM, by default all processor sets belong to pset_default where pool_default is associated with.
When user-defined pools and psets are created and containers are associated to these, its wise to make sure that pset_default contains
CPUs that belong to pyhsically separated boards. That is easily achieved with poolcfg.
Theoratically, this should keep the system running if a CPU goes offline, unless the Kernel panics for whatever reason due to a kernel function running at the time of crash,
then the result is a total system shutdown which may come up on the rest of CPUs on the other loosely-coupled boards after the reboot.
In this case, the system won't stay down until the h/w is replaced, but would run on the rest of CPUs on the other boards after the reboot.
This is all theoretical and needs to be proven !
Kimberly Chang wrote:
On a system with multiple containers and resource pools, what happens if the CPU's in the default pool that bound to the global zone become unavailable? Does the system stay down until the h/w (e.g a CPU board) is replaced?Best regards, Kimberly _______________________________________________ zones-discuss mailing list firstname.lastname@example.org
_______________________________________________ zones-discuss mailing list email@example.com