High CPU on LXD Host machine

LXD
1

Hi
I have created 6 containers on the LXD host machine and configured the limits of CPU/Memory. So, they can not impact other environments. However, they impacted the host server(other containers) and the CPU load goes crazy “load average: 850, 1012, 1090”

LXC version: 3.0.3

Container configuration:
architecture: x86_64
config:
image.architecture: amd64
image.description: Ubuntu bionic amd64 (20210323_07:42)
image.os: Ubuntu
image.release: bionic
image.serial: “20210323_07:42”
limits.cpu: “16”
limits.cpu.priority: “0”
limits.memory: 16GB
limits.memory.swap: “false”
raw.lxc: |
lxc.cgroup.devices.allow = a
lxc.cap.drop =
security.nesting: “true”
security.privileged: “true”
volatile.base_image: 1ee96c1ad98b6xxxx9542dc94cee568ad300e34ccfa322bc5xxxe77709adaa
volatile.eth0.hwaddr: 00:16:3e:1f:a1:9c
volatile.eth0.name: eth0
volatile.idmap.base: “0”
volatile.idmap.next: ‘[]’
volatile.last_state.idmap: ‘[]’
volatile.last_state.power: RUNNING
devices:
eth0:
ipv4.address: 10.111.222.15
nictype: bridged
parent: lxdbr0
type: nic
ephemeral: false
profiles:

  • default
    stateful: false
    description: “”

Am I missing anything to configure? Please help

Thanks & Regards,
Mangesh

2

Hi Mangesh_Dhamale,

One cause of this i have found is diskio which is caused by anything which is write intensive. Basically the CPU has to wait for data to be written to disk. From the host machine, you should be able to see this from running something like top or sysstat. Top will also give you an idea if it’s a process/container causing it.

I suspect that because the load goes through the roof, it is very tricky to debug and access the server to get any information from it.

I hope this is somehow helpful.