Open Stack

On Mon, 2019-11-11 at 14:44 -0800, Al Lau wrote:
In "/opt/kata/bin/kata-runime update --help", it has the "--cpuset-mems value" option.  I think this works if there is already a kata container running.  Is there a way to set the cpuset-mems to 0 in the configurations.toml file (for example) so that all the kata containers uses the NUMA node 0 on memory usage?

no, the number of nodes is not configurable from configuration file


In my environment, the kata containers are started through CRI API.  If not possible to preset cpuset-mems for all kata containers, I have to find a programmatic way to set the cpuset-mems.

since you are using ctr, you can use a custom runtime config

ctr run --help
--config value, -c value  path to the runtime-specific spec config file


Thanks,
Al

On Mon, Nov 11, 2019 at 9:56 AM Montes, Julio <julio.montes at intel.com<mailto:julio.montes at intel.com>> wrote:
they are used to specify in which nodes you allow the execution of the containers, these nodes contains information such as what CPUs are local to the node,
I'm not an expert on this, but I guess they can be used to reduce latency or improve something

On Mon, 2019-11-11 at 09:45 -0800, Al Lau wrote:
The nodes in /sys/devices/system/node are memory nodes and they are used for execution of the containers.  Is this the purpose of the nodes?

Thanks,
Al

On Mon, Nov 11, 2019 at 8:42 AM Montes, Julio <julio.montes at intel.com<mailto:julio.montes at intel.com>> wrote:
Thanks,

seems like by default the container engine that you're using tries to assign both nodes (0,1) to the container, unfortunately inside the VM (where the container is created)
there is only one node (0), as a temporal workaround, can you change this behavior? (i.e docker run --cpuset-mems="0" ....)


On Mon, 2019-11-11 at 08:36 -0800, Al Lau wrote:
$ ls -l /sys/devices/system/node
total 0
-r--r--r-- 1 root root 4096 Nov 11 08:32 has_cpu
-r--r--r-- 1 root root 4096 Nov 11 08:32 has_memory
-r--r--r-- 1 root root 4096 Nov 11 08:32 has_normal_memory
drwxr-xr-x 4 root root    0 Nov  8 01:45 node0
drwxr-xr-x 4 root root    0 Nov  8 01:45 node1
-r--r--r-- 1 root root 4096 Nov 11 08:32 online
-r--r--r-- 1 root root 4096 Nov 11 08:32 possible
drwxr-xr-x 2 root root    0 Nov 11 08:32 power
-rw-r--r-- 1 root root 4096 Nov  8 01:45 uevent
$ ls -l /sys/devices/system/node | grep node | wc -l
2

On Mon, Nov 11, 2019 at 6:28 AM Montes, Julio <julio.montes at intel.com<mailto:julio.montes at intel.com>> wrote:
On Mon, 2019-11-11 at 08:23 -0600, Julio Montes wrote:
Hi Al

how many memory nodes have your system? ( ls /sys/devices/system/ | grep node   | wc -l )

*  ls /sys/devices/system/node | grep node   | wc -l


can you change the number of cpuset-mems to 0? for example in docker: docker run --cpuset-mems="0" ....

-
Julio

On Fri, 2019-11-08 at 14:56 -0800, Al Lau wrote:
Hi Julio,
Attached is the configuration.toml file.  I hope this is what you are looking. for.
Thanks,
Al

On Fri, Nov 8, 2019 at 2:11 PM Montes, Julio <julio.montes at intel.com<mailto:julio.montes at intel.com>> wrote:
I'd like to see the kata configuration and pod configuration (yaml)

running custom kernel?

On Fri, 2019-11-08 at 14:05 -0800, Al Lau wrote:
My environment configured containerd with kata-runtime.  I followed the setup instructions on https://github.com/kata-containers/documentation/blob/master/how-to/containerd-kata.md#configuration

When I attempt to create a kata container, I got this error message.  Anyone know how I can resolve this?

rpc error: code = Unknown desc = failed to create containerd task: Could not run process: container_linux.go:346: starting container process caused "process_linux.go:297: applying cgroup configuration for process caused \"failed to write 0,1 to cpuset.mems: write /sys/fs/cgroup/cpuset/k8s.io/f1fdd8d5be9030388e4e04dd7d42df9e77a3d46ecb4922ecb2936bf92f74595d/cpuset.mems<http://k8s.io/f1fdd8d5be9030388e4e04dd7d42df9e77a3d46ecb4922ecb2936bf92f74595d/cpuset.mems>: numerical result out of range\"": unknown

kata-runtime  : 1.8.2
   commit   : cdc30296c370414d8ef445abd06378199775d076
   OCI specs: 1.0.1-dev

containerd github.com/containerd/containerd<http://github.com/containerd/containerd> 1.2.4 e6b3f5632f50dbc4e9cb6288d911bf4f5e95b18e

$ cat /etc/containerd/config.toml
. . .
        [plugins.cri.containerd.runtimes.kata]
          runtime_type = "io.containerd.kata.v2"
          [plugins.cri.containerd.runtimes.kata.options]
            ConfigPath = "/etc/kata-containers/configuration.toml"
      [plugins.cri.cni]
        bin_dir = "/opt/cni/bin"
        conf_dir = "/etc/cni/net.d"
        conf_template = ""

Thanks,
Al


_______________________________________________

kata-dev mailing list

<mailto:kata-dev at lists.katacontainers.io>

kata-dev at lists.katacontainers.io

<http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev>

http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.katacontainers.io/pipermail/kata-dev/attachments/20191112/6e781629/attachment.html>