Has anyone experienced the following in their setup:
1.) All NF’s run in k8s/docker --> Latest stable version
2.) CNI Plugin used Cilium (Used Flannel initially but flakey)
3.) Most NF k8s services are ClusterIP with NRF, AMF, UPF’s all configured with nodePort
3.) Once I reboot the Master Node the UPF’s go into CrashLoopBackOff
- They never get to the point where the static IP addresses are allocated for a simulated network
- Biggest complaint seems to be a /dev/mqueue full which doesn’t reflect correctly the case
- I wrote small little script for now which is messy to delete the UPF pods, recreate
- Everytime this messy fix works and am wondering is it ore of a timing issue within k8s ?
Any thoughts of ideas wlecomed
Donal