-
Bug
-
Resolution: Not a Bug
-
Normal
-
None
-
None
-
None
-
False
-
-
False
-
-
-
uShift Sprint 234
Description of problem:
Enabling microshift on aarm fedora coreos leads to endless restart loop of microshift service. It panics while creating SCCs. Error is nil pointer reference.
656 │ I0325 11:22:28.005925 7014 openshift-oauth-server.go:112] oauth-apiserver is ready
657 │ I0325 11:22:28.006648 7014 manager.go:112] Starting openshift-default-scc-manager
658 │ I0325 11:22:28.009248 7014 scc.go:73] Applying scc api assets/scc/0000_20_kube-apiserver-operator_00_scc-anyuid.yaml
659 │ E0325 11:22:28.033401 7014 runtime.go:76] Observed a panic: runtime error: invalid memory address or nil pointer dereference
660 │ goroutine 26078 [running]:
661 │ k8s.io/apiserver/pkg/server/filters.(*timeoutHandler).ServeHTTP.func1.1()
662 │ /src/microshift/packaging/rpm/_rpmbuild/BUILD/microshift-48863ff4cf9146906a7d7879c2ca93265c7ad662/vendor/k8s.io/apiserver/pkg/server/filt
│ ers/timeout.go:102 +0xb0
663 │ panic({0x589f600, 0x90838c0})
664 │ /usr/lib/golang/src/runtime/panic.go:884 +0x20c
665 │ reflect.mapiterinit(0x57cd380?, 0x401ea09c78?, 0x401ea21bf8?)
666 │ /usr/lib/golang/src/runtime/map.go:1375 +0x1c
667 │ github.com/modern-go/reflect2.(*UnsafeMapType).UnsafeIterate(...)
Version-Release number of selected component (if applicable):
How reproducible:
Steps to Reproduce:
I installed fedora coreos on raspi 4 and compiled rpms for microshift by following these steps (I built it in fedora 37 container on the coreos on raspi4):
https://microshift.io/docs/developer-documentation/build-install-rpm/
and installed it by following these:
https://microshift.io/docs/getting-started/
Then after installing crio, sellinux and microshift rpm:
- systemctl start microshift
Actual results:
It proceeds to raise the etcd and api up, but then while doing SCCs it panics. See the full log in attachment.
Expected results:
Start the services and start serving API.
Additional info:
If one has raspi, here are some details of the image:
config.bu
variant: fcos
version: 1.4.0
passwd:
users:
- name: core
ssh_authorized_keys:
- <ssh key>
storage:
links:
- path: /etc/localtime
target: ../usr/share/zoneinfo/Europe/Helsinki
files:
- path: /etc/hostname
overwrite: true
contents:
inline: raspishift
disks:
- device: /dev/disk/by-id/coreos-boot-disk
wipe_table: false
partitions:
- label: root
number: 4
size_mib: 51200
wipe_partition_entry: true
- label: swap
number: 5
size_mib: 8192
guid: 0657FD6D-A4AB-43C4-84E5-0933C84B4F4F
wipe_partition_entry: true
- label: data
number: 6
guid: E6D6D379-F507-44C2-A23C-238F2A3DF928
wipe_partition_entry: true
and creating the install image for coreos from above:
butane --pretty --strict config-rootfs.bu|tee butane.ign sudo coreos-installer install --architecture=aarch64 -i butane.ign /dev/sda
FCOSEFIPARTITION=$(lsblk $FCOSDISK -J -oLABEL,PATH | jq -r '.blockdevices[] | select(.label == "EFI-SYSTEM")'.path)
mkdir /tmp/FCOSEFIpart
sudo mount $FCOSEFIPARTITION /tmp/FCOSEFIpart
sudo rsync -avh --ignore-existing /tmp/RPi4boot/boot/efi/ /tmp/FCOSEFIpart/
sudo umount $FCOSEFIPARTITION
that gives a usb drive which will boot raspberyrpi into fedora coreos. Then I grabbed fedora container to build the rpm in:
podman run -ti --name fed --hostname fed -v src:/src:z fedora:37
I build the rpm in there according to above link for instructions, and installed the rpms to the fcos.
