no space left on device after a fresh installation

[zero-universe]

Hi,

I just installed 3 talos 1.7.0 nodes (1 controlplane, 2 workers) on a 100GB drive each.
Then I deployed k8s gateway api crds, metallb, envoy gateway and cert-manager.

When I apply a simple deployment with a nginx pod and one replica,
my deployment complains with "no space left on device".

All 6 partitions were created on each host (https://www.talos.dev/v1.7/learn-more/architecture/),
and talosctl usage shows at least 80GB free on the "EPHEMERAL" partition.

Did I miss something ?

thanks

[smira]

If you have a problem, please submit full logs. What component complains, how does it complain exactly?

Also please attach the talosctl support output.

[zero-universe]

is there a command like:

• lsblk
• df
• findmnt
• cat /proc/partitions

[zero-universe]

this is my deployment:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: void-nginx
  namespace: dreamingvoid
spec:
  selector:
    matchLabels:
      run: void-nginx
  replicas: 2
  template:
    metadata:
      labels:
        run: void-nginx
    spec:
      containers:
      - name: void-nginx
        image: nginx
        ports:
        - containerPort: 80
        securityContext:
          allowPrivilegeEscalation: false
          runAsNonRoot: false
          capabilities:
            drop:
              - ALL
          seccompProfile:
            type: RuntimeDefault
        resources:
          requests:
            cpu: 80m
            memory: 256
          limits:
            cpu: 200m
            memory: 512

and the culprit seems to be the "resources" part.
without it, there are no problems ...

[smira]

sorry, I missed the point here, do you request more ephemeral than available?

[zero-universe]

hi,
no I did not.

        resources:
          requests:
            cpu: 80m
            memory: 256
          limits:
            cpu: 200m
            memory: 512

commenting this out, did the trick.

[zero-universe]

there are no limits defined, so i just wanted to set some for cpu and ram, but it seems not to work in my case.

[zero-universe]

all nodes are VMs running on KVM, hard drives are virtio devices (/dev/vda).