github: add workflows for building and publishing kata artefacts on ppc64le #8459
Conversation
|
Hi @fidencio @stevenhorsman , can you please help me in adding a self hosted ppc64le runner? |
Sure, send me a slack when you are ready and I can send through the token and instructions |
- statically linked qemu requires slof.bin to run, hence remove it from blacklist - By default, initrd is used for Power, modify the configuration.toml accordingly Fixes: kata-containers#8458 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
Since the release is not yet out for ppc64le, build oras from source and use it. Fixes: kata-containers#8458 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
Amulyam24
force-pushed
the
workflow-1
branch
from
5e953dd
to
0c894b0
Compare
This PR adds workflow to run containerd tests on Power as a part of CI migration. Fixes: kata-containers#8500 Depends-on: kata-containers#8459 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
This PR adds workflow to run containerd tests on Power as a part of CI migration. Fixes: kata-containers#8500 Depends-on: kata-containers#8459 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
There was a problem hiding this comment.
Thanks for the request. I've left 2 comments.
Additionally, I've noticed that we finally have workflows to build artifacts for 4 different architectures where no steps including a if: always() condition for a self-hosted runner (running 24/7) are engaged, but to produce lots of duplicated codes.
I was wondering if this PR would be a right place to incorporate the workflows or we could do it in a separate PR. @fidencio and @stevenhorsman
Amulyam24
force-pushed
the
workflow-1
branch
from
0c894b0
to
3aa6a69
Compare
There was a problem hiding this comment.
Other than the comment regarding the deduplication, LGTM! Thanks @Amulyam24 !
This PR adds workflow to run containerd tests on Power as a part of CI migration. Fixes: kata-containers#8500 Depends-on: kata-containers#8459 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
…pc64le Adds workflows for building kata static tarball and releasing it. Fixes: kata-containers#8458 Signed-off-by: Amulyam24 <amulmek1@in.ibm.com>
Amulyam24
force-pushed
the
workflow-1
branch
from
3aa6a69
to
ae2c0c5
Compare
There was a problem hiding this comment.
It would be good to see a run on this testing on your fork, but I think the code looks okay and expected.
I definitely understand @BbolroC's comments about the duplication:
- The build-kata-static-tarball workflow only differs in artifact name, runner and assets that are created
- The publish workflows only varies on the artifact name and self-host runner prepare in ppc
- The release workflows are pretty identical other than the build workflows they call
I'd like to hear @fidencio's thought, but I think we should consider refactoring these to try and get a single workflow with passing the arch in. We might have a couple of steps like:
- name: Prepare the self-hosted runner
if: ${{ arch == 'ppc64le' }}
run: ${HOME}/scripts/prepare_runner.sh)
but I think it's worth tying to see if it's not too much of a mess and would potentially save on the maintenance cost of updating 4 versions in future.
I'm okay for this to be done as a separate PR though.
|
@stevenhorsman, testing done for both this and containerd PR, PTAL for reference - https://github.com/Amulyam24/kata-containers/actions/runs/6979218191/job/18992074251?pr=6 |
|
Hi @stevenhorsman, we have 3 approvals, can this PR be merged? |
|
/test |
I am with Steve and Choi here, but I'd for sure prefer this addition coming on a different PR. |
This is the first step of the migration of jobs from Jenkins to GHA.
Fixes: #8458