Re: [libvirt-users] /proc/meminfo

Hi all yes, it's occured time to time on our installations. Centos 7.2 + libvirt 1.2.18 and probably on 1.3.2 We have workaround for fix it without reboot LXC container. 1) check that on HW node exists cgroups memory for container. [root@node]# cat /sys/fs/cgroup/memory/machine.slice/machine-lxc\\x2dpuppet.infra.scope/memory.limit_in_bytes 17179869184 [root@node]# cat /sys/fs/cgroup/memory/machine.slice/machine-lxc\\x2dpuppet.infra.scope/memory.memsw.limit_in_bytes 18203869184 In our case limit exists and set to 16Gb mem and 16+1 Gb for mem +swap Contaner name puppet.infra, substitute here your container name. 2) if exists - simple attach cgroups to libvirt_lxc pid : node# yum install libcgroup-tools node# cgclassify -g memory:machine.slice/machine-lxc\\x2dLXC_CONTAINER_NAME.scope PID where pid in my case found as: [root@node]# ps ax | grep libvirt_lxc | grep -v grep | grep puppet 22254 ? Sl 296:25 /usr/libexec/libvirt_lxc --name puppet.infra --console 24 --security=none --handshake 42 --veth macvlan0 --veth macvlan1 After run cgclassify /proc/memory inside container show normal values In some casees, on combination old libvirt 1.2.xx and old systemd in some conditions (after systemctl restart libvirtd or daemon-reload) we get same error. But in such case we need first create/restore cgroups on node and after this cgclassify process to groups. For example: make dir # mkdir "/sys/fs/cgroup/memory/machine.slice/machine-lxc\x2dpuppet.scope" set limit: [root@]# echo 8589934592 > /sys/fs/cgroup/memory/machine.slice/machine-lxc\\x2dpuppet.scope/memory.limit_in_bytes [root@]# echo 9663676416 > /sys/fs/cgroup/memory/machine.slice/machine-lxc\\x2dpuppet.scope/memory.memsw.limit_in_bytes classify: #cgclassify -g memory:machine.slice/machine-lxc\\x2dpuppet.scope PID_LIBVIRT_LXC p.s. use libvirt 1.3.2, it's more stable and never show negative values for memory and swap inside container b.r. Maxim Kozin