[hotfix] Synchronize CI pipeline setup #78
Conversation
.github/workflows/push_pr.yml
Outdated
| strategy: | ||
| matrix: | ||
| flink: [ 1.17.1, 1.18.0 ] | ||
| uses: apache/flink-connector-shared-utils/.github/workflows/python_ci.yml@ci_utils | ||
| with: | ||
| flink_version: ${{ matrix.flink }} |
There was a problem hiding this comment.
should we remove ci for python part?
IIRC currently it is done separately
There was a problem hiding this comment.
Oh wow I completely missed the Python part. Let me check
There was a problem hiding this comment.
I think we should keep them in for PRs, and actually we should also have them for weekly runs.
|
since it is touched a couple of more ideas
|
The Flink Kafka connector now tests: - All PRs against `main` are tested against Flink `1.17.2` and `1.18.1`, instead of `1.17.1` and `1.18.0` - The weekly run tests -- `main` against `1.17-SNAPSHOT, `1.18-SNAPSHOT` and `1.19-SNAPSHOT` (1.19 is newly added) -- `v3.0` against `1.17-SNAPSHOT` and `1.18-SNAPSHOT` (as was before)
MartijnVisser
force-pushed
the
sync_ci_1.19
branch
from
3a66a05 to
070127f
Compare
I think we should mirror what we do with Flink, so I think we should indeed also run them for JDK11 and JDK17? We could argue that JDK11/JDK17 would only be run during a weekly, to avoid overly hitting the servers. I believe we also run those tests during the nightly builds for Flink
We should indeed |
|
yeah... initially there were already 2 jdks (8, 11) in ci for PRs and that's why I've just added other jdks there... it would make sense to follow something similar we have in Flink main repo where everything is built with jdk8 for PRs and all other jdks are participating only during nightlies |
The Flink Kafka connector now tests:
mainare tested against Flink1.17.2and1.18.1, instead of1.17.1and1.18.0mainagainst1.17-SNAPSHOT,1.18-SNAPSHOTand1.19-SNAPSHOT(1.19 is newly added) --v3.0against1.17-SNAPSHOTand1.18-SNAPSHOT` (as was before)