On 12/20/2016 at 10:24 AM, Pratyush Anand wrote:
Hi Xunlei,
Thanks for the initiative on it.
On Monday 19 December 2016 08:54 PM, Xunlei Pang wrote:
> e.g. some x86_64 large machine : 24TB RAM, nearly 1024 cpus, with the default
> kdump configuration, kdump kernel hung at boot up stage. Manually changed to
> "crashkernel=512M,high crashkernel=256M,low nr_cpus=4", then it works
properly.
So, if we have only 1G of RAM and all other configuration as above, then will that
machine boot with nr_cpus=2?
Yes, if it can pass the x86_64 vector special treatment, the result will be 2 under this
patchset.
I mean, if there is some other factor than memory which governed minimum number of CPUs
then automating nr_cpus on the basis of memory might fail in some cases.
currently the insufficient kdump memory issue and vector issue(for x86_64) are the two
factors we've met.
But there may be others lurking somewhere.
Regards,
Xunlei