malinjawi opened a new pull request, #13370: URL: https://github.com/apache/gluten/pull/13370
## What changes are proposed in this pull request? CI jobs that run inside the `apache/gluten:*` images fail during setup when the EPEL mirrors are out of sync, even if none of the packages being installed come from EPEL. Example: apache/gluten#13357, job `spark-test-spark41 (1)` (run 39123817210). Its step `dnf install -y python3.11 python3.11-pip python3.11-devel` failed with `Failed to download metadata for repo 'epel'` after every mirror returned 404. python3.11 comes from AppStream, but dnf aborts when any enabled repo has `skip_if_unavailable=False` (the default for the epel repo files). This PR sets `skip_if_unavailable=True` on the `epel*` repo files in the image Dockerfiles right after `epel-release` is installed. It is one small edit per Dockerfile, not an edit to the ~89 install lines in the workflows. It does not mask real failures. With EPEL unreachable, installing an EPEL-only package (e.g. `patchelf`, `ccache`, `gh`) still fails with `No match for argument`; only unrelated installs stop failing. BaseOS/AppStream/CRB are left strict. Frequency (apache/gluten, 2026-09-09..2026-10-19, 717 failed non-lint runs, 2090 failed jobs scanned): 25 jobs failed on repo metadata or RPM mirror errors. 7 of them were EPEL (`epel` x6, `epel-next` x1): 2026-10-18, 10-14 (x2), 10-10, 10-08 (x2), 10-07. The rest were BaseOS, AppStream, Stream-AppStream, openEuler or other mirror errors, which this PR intentionally does not address. Takes effect only after the images are rebuilt and pushed by the `Build and Push Docker Image` workflow (weekly cron or a push to main touching `dev/docker/*`). ## How was this patch tested? Ran the same `sed` in the `quay.io/centos/centos:stream9`, `almalinux:8` and `centos:8` (vault) base images after installing `epel-release`: all `epel*` repo sections get `skip_if_unavailable=True`. On `apache/gluten:centos-9-jdk17` with the EPEL baseurl pointed at a dead host: before, `dnf install bison` fails with the exact error above; after, it succeeds, and `dnf install patchelf` still fails with `No match for argument`. The full image builds were not run locally. ## Was this patch authored or co-authored using generative AI tooling? Generated-by: IBM Bob -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
