On 10/30/2013 09:37 AM, Peter Sabaini wrote:
On 30.10.13 16:21, Kir Kolyshkin wrote:
On 10/30/2013 07:34 AM, Peter Sabaini wrote:
Hey all,
I've read somewhere that with OpenVZ kernel > 2.6.32-042stab072.10 it
should be possible to set sysctl from within the VE
However:
# sysctl -w fs.file-max=10000
error: permission denied on key 'fs.file-max'
# uname -a
Linux xxx 2.6.32-042stab081.5 #1 SMP Wed Oct 23 16:32:28 MSK 2013
x86_64 x86_64 x86_64 GNU/Linux
You have to change this on the host system (i.e. hardware node).
For container, this is limited by numfile UBC
(http://openvz.org/numfile)
Yeah...
Maybe I should elaborate on my usecase: I'm testing my Salt
(http://saltstack.com/) configuration (which ultimately deploys to
metal) inside OpenVZ.
Works great, I especially appreciate the fact that a new VE can be
quickly built/restarted and I can have a simulation of a whole cluster
in a VMware box thanks to the low overhead of VEs.
However, sometimes I would like to do low-level config via Salt as
well, and up to now I've relied on ugly hacks to make Salt think that
# sysctl -p etc. have actually succeeded and I was wondering if that
was still necessary.
So from what you're saying I gather that that hasn't changed and I
can't write kernel sysctls from VE, right?
There are some sysctls you can change from a container, I'm just not
sure what the list is.
Most of them, though, are read-only. Please file a kernel bug with
detailed description of what you need and why.
_______________________________________________
Users mailing list
Users@openvz.org
https://lists.openvz.org/mailman/listinfo/users
