dragonflyoss/Dragonfly2

No traces when pulling images from ACK containerd

Opened this issue · 5 comments

Bug report:

k8s version is 1.26, containerd version is 1.6.20, dragonfly helm package verison is 1.1.29

  1. The Dragonfly has been deployed on Alibaba Cloud ACK and two registries have been set up
  containerd:
    enable: true
    injectConfigPath: true
    registries:
      - "https://ghcr.io"
      - "https://lthub-registry.cn-beijing.cr.aliyuncs.com"
  1. All services have started logging normally, referencing dfdaemon.
dfdaemon:
  enable: true
  config:
    console: true
    verbose: true
    pprofPort: 18066
    jaeger: ""
jaeger:
  enable: true
  1. The storage is configured with 20Gi to comply with the minimum quota requirement of Alibaba Cloud's cloud disk.
  2. Confirm that all local services have been successfully started.
    image
  3. Look at backend services, Confirm All Are Accessible Normally
    image
    image
  4. Based on official use cases for testing, corresponding links https://d7y.io/docs/getting-started/quick-start/kubernetes/#containerd-pull-image-back-to-source-for-the-first-time-through-dragonfly
    image
  5. Check that the configuration for containerd on the instance is generated correctly
    image
  6. The correct request records for the official website use case were not found on Jaeger.
    image

Expected behavior:

image

How to reproduce it:

Using Helm to Deploy on ACK Cluster

Environment:

  • Dragonfly version: 2.1.30
  • OS: 1.26.3-aliyun.1
  • Kernel (e.g. uname -a): Linux testompub-pub-172.17.29.36static 5.10.112-11.2.al8.x86_64 #1 SMP Mon Aug 1 20:06:08 CST 2022 x86_64 x86_64 x86_64 GNU/Linux
  • Others:
    • containerd: 1.6.20
    • kubelet: Kubernetes v1.26.3-aliyun.1

one dfdaemon console log file
dfdaemon.log
What logs do I need to provide?

Check whether the config_path in the containerd configuration is correct, refer to https://github.com/containerd/containerd/blob/main/docs/cri/config.md#registry-configuration.

If correct, please check the containerd log. Why is mirror not used?

Check whether the config_path in the containerd configuration is correct, refer to https://github.com/containerd/containerd/blob/main/docs/cri/config.md#registry-configuration.

If correct, please check the containerd log. Why is mirror not used?

I found it!This is default config in ACK /etc/containerd/config.toml.
image
The default configuration of ACK is missing an "s"

What happens next is up to the community to handle the situation.
dragonfly/templates/dfdaemon/dfdaemon-daemonset.yaml
It is recommended to read config_path from the current node and not hardcode it in the deployment script.

What happens next is up to the community to handle the situation. dragonfly/templates/dfdaemon/dfdaemon-daemonset.yaml It is recommended to read config_path from the current node and not hardcode it in the deployment script.

@sigboom Agree!