Include fix for bcvk - #113
Conversation
We need a fix for unlocking our CI which requires the commit in: bootc-dev/bcvk#313 afterwards the commit inside the disk image should match with the one of the corresponding node image. Signed-off-by: Alice Frosi <afrosi@redhat.com>
Printing the digest of the node image should help us to validate if it is the same digest used by bootc inside the bcvk VM. Signed-off-by: Alice Frosi <afrosi@redhat.com>
e7a9a13 to
b4bf404
Compare
bcvk uses virtiofsd to expose the bootc OCI image's root filesystem to the ephemeral VM via virtio-fs. Without it the kernel boots past BIOS and GRUB but hangs waiting for its root device, so SSH never becomes available and the 240s timeout fires. virtiofsd is only a weak/recommended dependency of qemu-kvm, so --setopt=install_weak_deps=0 silently dropped it. Add it explicitly. Assisted-by: AI Signed-off-by: Alice Frosi <afrosi@redhat.com>
bcvk's console log capture reads from the virtio console (hvc0), but the kernel cmdline only listed tty0 and ttyS0. Assisted-by: AI Signed-off-by: Alice Frosi <afrosi@redhat.com>
bcvk generates an SSH keypair via ssh-keygen before starting QEMU. The openssl-client was missing. Assisted-by: AI Signed-off-by: Alice Frosi <afrosi@redhat.com>
00c29c5 to
b863f2a
Compare
|
Bink CI is pretty flaky unfortunately, but I do really want to get this merged since it fixed bootc-operator ci. We can investigate the CI stability in a later PR |
|
So, I tested bootc-dev/bcvk#313 locally and it didn't seem to solve the issue as the manifest changes with oci dir too. I have bootc-dev/bcvk#314 which might fix it |
| @@ -1,7 +1,16 @@ | |||
| ARG FEDORA_VERSION=44 | |||
| ARG BCVK_REF=main | |||
There was a problem hiding this comment.
Containerfile has ARG BCVK_REF=main but the Makefile pins BCVK_REF ?= 9005...
I think it will make difference when someone uses makefile or podman build. like they will have different thing pinned/unpinned
There was a problem hiding this comment.
if you use podman build you should know what to put as build argument otherwise, it will take it from main which is fine IMO
|
@Johan-Liebert1 it does seems to fix the ostree one since the digest matches. During the installation for composefs, I'm not able to see any reported digest. This is for sure an improvement we could also implement |
|
I'm going to merge this as it seems fixing the ostree build and it should unblock the bootc-operator CI. It will help also when the fix for composefs lands since we now have the ability to build bcvk from source |
3 participants
Include fix in bcvk: bootc-dev/bcvk#313. Also, install virtiofs and openssl dependencies and enable hvc0 console.
This should hopefully fix bootc-dev/bootc-operator#136 . Since now the digest printed by the node images matches with the one use by bootc into the bcvk VM