CHR: intermittent single-core 100% CPU usage in cloud and management

RouterOS Virtualization
1

Hello,

I am experiencing an intermittent high-CPU issue on RouterOS CHR.

This issue has occurred on both RouterOS 7.21.4 and 7.21.5 long-term, so it does not appear to be specific to only one of these versions.

After the CHR instance has been running normally for some time, one of the two vCPUs may become continuously utilized at approximately 96–100%. The condition does not recover automatically, even after network traffic returns to a low level.

Environment:

  • Platform: CHR x86_64
  • RouterOS versions affected: 7.21.4 and 7.21.5 long-term
  • Hypervisor: Proxmox VE 8.4
  • VM machine type: Q35
  • Host CPU: Intel N4100
  • VM resources: 2 vCPUs, 512 MiB RAM
  • License: P10
  • WAN: PPPoE
  • NICs: Intel I226-V
  • WAN and LAN NICs are passed through directly to the CHR VM using PCI passthrough

When the issue occurs, /system resource print shows approximately 50% total CPU usage, which is consistent with one of the two vCPUs being fully utilized.

The profiler shows that the load is mainly attributed to cloud and management.

Example:

management    cpu0    41.5%
cloud         cpu0    54%
cpu0                  96%

At another time, the same load appeared on CPU1:

management    cpu1    34%
cloud         cpu1    59.5%
cpu1                  96%

The high load can move between CPU0 and CPU1, so it does not appear to be permanently tied to one specific core or one fixed NIC interrupt.

Additional observations:

  • winbox usage is around 0–0.5%
  • firewall usage is around 0–0.5%
  • networking, ethernet, bridging, and igc are close to 0%
  • IPv4 and IPv6 connection counts are low
  • Current traffic can be very low while the CPU core remains fully utilized
  • No custom script job was running when the issue was checked
  • Configured Scheduler tasks had not run since boot
  • The CHR license status appears normal
  • IP Cloud configuration is:
ddns-enabled: auto
ddns-update-interval: none
update-time: no

The issue is intermittent and I do not yet have reliable reproduction steps.

During the most recent occurrence, I had run an Internet speed test from a LAN client before noticing the problem. However, this should only be treated as an observation, not as a confirmed trigger.

Most previous occurrences happened while I was away from home, so I could not determine what traffic or activity occurred immediately before the CPU usage increased.

I generated a supout.rif file while the problem was active and will only provide it privately to MikroTik Support. I will not upload it publicly to the forum.

Has anyone else seen similar behavior on RouterOS CHR, especially with Proxmox VE and PCI-passthrough Intel I226-V NICs?

I would also like to know whether:

  • this is a known CHR issue;
  • there is a workaround;
  • a testing or development build contains a related fix;
  • any additional logging should be enabled before the issue occurs again.

Thank you.

2

Hi,

Have you checked forum for that problem? There are some topics on that.

image
image1257×1167 173 KB

3

Yes, I have already investigated the issue with the help of AI and also searched the MikroTik forum.

I found one forum topic where the reported behavior was very similar to mine. The main conclusion there was to avoid using PCI passthrough for the network adapters.

I have also tested the CHR VM without NIC passthrough. However, the performance of my device is relatively limited, and without PCI passthrough the router cannot reach the full speed of my Internet connection.

Therefore, avoiding NIC passthrough is not a practical long-term workaround in my environment.