kubevirtci/cluster-up/kubectl.sh version
will try to connect to the apiserver.
To investigate the logs of a pod, you can view the logs via
kubevirtci/cluster-up/kubectl.sh logs
. To view the logs of virt-api
, type
kubevirtci/cluster-up/kubectl.sh logs virt-api -f
Sometimes a container in a pod is crashlooping because of an application error
inside it. In this case, you normally can't see any logs, because the container
is already gone, and so are the logs. To get the logs from the last run
attempt, the --previous
flag can be used. To view the logs of the container
virt-api
in the pod virt-api
from the previous run, type
kubevirtci/cluster-up/kubectl.sh logs virt-api -f -c virt-api -p
Note that you always have to select a container inside a pod for fetching old
logs with the --previous
flag.
Both, Kubernetes and KubeVirt are creating events, which can be viewed via
kubevirtci/cluster-up/kubectl.sh get events --all-namespaces --watch
This way it is pretty easy to detect if a Pod or a VMI got started.
It can be very valuable to enter a container and do some investigations there,
to see what is going wrong. In this case the kubectl exec
command can be
used. To enter virt-api
with an interactive shell, type
kubevirtci/cluster-up/kubectl.sh exec virt-api -c virt-api -i -t -- sh
After all you might not see errors in the logs provided by Kubernetes. In that case
you can take a look at the logs of the kubelet
on the host where the issue is
appearing. Depending on the error it is getting logged to either the system logs or
to the kubelet logs, you can use the following commands to view them:
journalctl
# or
journalctl -u kubelet
This shows the basic principle on how remote debugging can be done.
- Add delve to the container
- Start delve on a specific port (
dlv attach <pid> --headless=true --api-version=2 --listen=:1234
) - Use kube-proxy to forward the port to your machine
A go debugger can be attached to kubevirt processes in the following manners:
Not all processses can easily be executed locally, since some of the processes require specific
dependencies that exist in the runtime pod and node
For processes that are easily executed locally, such as virt-controler
, the following program arguments
can be passed to virt-controller --kubeconfig /path/to/kubeconfig --leader-elect-lease-duration 99h
in order
to successfully debug. Since virt-contoller and other control plane components rely on leader election to achive
consensus among multiple replicas, and we only want our local replica to be the leader, even in the absence of other
replicas, we can temporarily grant control to our local replica by specifying the --leader-elect-lease-duration
argument, as shown above, with an example value of 99h
, granting the lease for 99 hours.
Note
/path/to/kubeconfig
must point to a running kubrenetes cluster with kubevirt installedNote It is recommended to scale down both
virt-controller
and alsovirt-operator
deployments to 0 replicas, because otherwise it will override thevirt-controller
deployment manifest.
Remote debugging may be more complex to setup, but has the advantage of executing the procesess in its real runtime environment, where all dependencies ae satisfied, therefore it can work for any go process
Compile the required go cmd with -gcflags='all=-N -l'
to produce
a debuggable binary.
Validate that it is in fact debuggable by running file <binary file>
and expect to see with debug_info, not stripped
in the output.
For Bazel
based build, add --@io_bazel_rules_go//go/config:gc_goopts=-N,-l --strip=never
to the bazel
build command.
Build a container image, with the compiled binary and publish to a registry that is accessible to the cluster
consider adding dlv
executable to the image for step 2
Having scaled down virt-operator, patch or edit the respetive k8s workload manifest
(e.g. virt-controller
Deployment, virt-handler
Daemonset ) to:
- Consume the debug image
- Execute
dlv
, either in a separate container, and attach to the debugged process as shown above. kubectl debug can be used, to easily add another containr to the pod. Note thatshareProcessNamespace: true
is necessary, for dlv to be able to debug a process running in another container. The other alternative is to modify the app containercmd
to havedlv
exec
the process in the same container:dlv --listen=:2345 --headless=true --api-version=2 --accept-multiclient exec <binary>
- Turn off probes, tweak securityContext and anything else that could interfere with slow execution
Here's a working example of a virt-controller
deployment patch
cluster-up/kubectl.sh --namespace kubevirt patch deployment virt-controller --type='json' -p '[
{
"op":"add",
"path":"/spec/template/spec/containers/-",
"value":{
"name":"dlv-debugger",
"image":"golang:1.23-alpine",
"command":[
"sh",
"-c",
"apk add --no-cache git bash && go install github.com/go-delve/delve/cmd/dlv@latest && \\
dlv attach $(pgrep virt-controller) --headless --accept-multiclient --api-version 2 --listen=:2345"
],
"securityContext":{
"seccompProfile":{
"type":"Unconfined"
},
"runAsUser":0
}
}
},
{
"op":"add",
"path":"/spec/template/spec/shareProcessNamespace",
"value":true
},
{
"op":"replace",
"path":"/spec/template/spec/containers/0/image",
"value":"registry:5000/kubevirt/virt-controller:debug"
},
{
"op":"replace",
"path":"/spec/replicas",
"value":1
},
{
"op":"replace",
"path":"/spec/template/spec/containers/0/imagePullPolicy",
"value":"Always"
},
{
"op":"replace",
"path":"/spec/template/spec/containers/0/securityContext",
"value":{
"runAsUser":0
}
},
{
"op":"replace",
"path":"/spec/template/spec/securityContext",
"value":{
}
},
{
"op":"remove",
"path":"/spec/template/spec/containers/0/readinessProbe"
},
{
"op":"remove",
"path":"/spec/template/spec/containers/0/livenessProbe"
}
]'
kubectl port-forward <pod> 2345