En la entrada Troubleshooting Kubernetes (II) encontré restos de la instalación de Flannel en la Raspberry Pi. Eliminé los pods que hacían referencia a Flannel y conseguí que el nodo k2 no se volviera a colgar.

Sin embargo, el problema sigue dándose en el nodo k3.

Revisando el contenido de /var/lib/kubernetes/pods/ he visto que algunos pods hacían referencia, todavía, a Flannel.

...
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~470691428.deleting~439642480.deleting~747926470.deleting~067946013.deleting~791070092.deleting~964331938.deleting~717873461.deleting~755129373.deleting~499171027
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~504683568.deleting~184413472.deleting~138413964.deleting~985160408.deleting~943143520.deleting~459558341.deleting~578589077.deleting~501462031.deleting~769373718
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~613861491.deleting~841547526.deleting~012178845.deleting~177797190.deleting~192052322.deleting~958792988.deleting~338401309.deleting~623810479.deleting~369130424
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~470691428.deleting~288881360.deleting~534630955.deleting~520377076.deleting~598159984.deleting~426698803.deleting~142931759.deleting~872800923.deleting~808586860
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~470691428.deleting~439642480.deleting~747926470.deleting~067946013.deleting~791070092.deleting~622848191.deleting~646325460.deleting~868409130.deleting~824166496
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~668636622.deleting~334825066.deleting~737147422.deleting~055159245.deleting~572255670.deleting~485248219.deleting~690855316.deleting~753094008.deleting~457647557
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~756292309.deleting~273222811.deleting~039503494.deleting~182629307.deleting~984614903.deleting~081831640.deleting~628560452.deleting~303652395.deleting~450650534
/var/lib/kubelet/pods/3a5e2819-21e5-11e7-bcfd-b827eb650fdb/volumes/kubernetes.io~configmap/wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_wrapped_flannel-cfg.deleting~883601122.deleting~535739903.deleting~385002935.deleting~558075878.deleting~174007749.deleting~757820208.deleting~194356513.deleting~813327027.deleting~485662152
...

Esta vez he detectado el problema al intentar calcular el espacio usado por esta carpeta, ya que la Raspberry Pi se ha quedado como “colgada”, aunque al lanzar htop no se observaba un uso excesivo de CPU.

Finalmente, he usado el mismo sistema que la otra vez: eliminar todas las subcarpetas de cada uno de los pods (dejando únicamente los que no se pueden borrar al estar en uso).

Después de la purga masiva de rm -rf /var/lib/kubelet/pods/ sólo han quedado dos carpetas en uso; el número corresponde con el número de pods planificados sobre el nodo k3 desde kubectl:

$ kubectl get pods --all-namespaces -o wide
NAMESPACE     NAME                         READY     STATUS    RESTARTS   AGE       IP             NODE
kube-system   etcd-k1                      1/1       Running   4          36d       192.168.1.11   k1
kube-system   kube-apiserver-k1            1/1       Running   4          36d       192.168.1.11   k1
kube-system   kube-controller-manager-k1   1/1       Running   4          36d       192.168.1.11   k1
kube-system   kube-dns-279829092-1b27r     3/3       Running   12         36d       10.32.0.2      k1
kube-system   kube-proxy-20t3b             1/1       Running   0          25m       192.168.1.13   k3
kube-system   kube-proxy-3dggd             1/1       Running   4          36d       192.168.1.11   k1
kube-system   kube-proxy-5b8k3             1/1       Running   2          12d       192.168.1.12   k2
kube-system   kube-scheduler-k1            1/1       Running   4          36d       192.168.1.11   k1
kube-system   weave-net-6qr0l              2/2       Running   8          36d       192.168.1.11   k1
kube-system   weave-net-mxp2w              2/2       Running   0          25m       192.168.1.13   k3
kube-system   weave-net-tmmdj              2/2       Running   4          12d       192.168.1.12   k2
$

Un reinicio y ¡listo!, problema -espero- resuelto de forma definitiva.