Prep for upgrade job infra fixups #22073
Conversation
|
Labelling this PR as size/L |
k8s-github-robot
added
the
size/L
|
GCE e2e build/test failed for commit d2dec87c5bef501977c9a1c1fac4e3a5ee3dd82b. Please reference the list of currently known flakes when examining this failure. If you request a re-test, you must reference the issue describing the flake. |
| else | ||
| # .config and its children are created by the gcloud call that we use to | ||
| # get the GCE service account. | ||
| # console-log.txt is created by Jenkins, but is usually not flushed out |
There was a problem hiding this comment.
Either here or in the near future we should remove the console-log.txt exception, since I think @ixdy got rid of the trap call in upload-to-gcs.sh.
There was a problem hiding this comment.
Nevermind; I got scared and put it back.
|
@spxtr PTAL. I'm happy to continue pulling things into functions if you'd like; let me know. I think this is minimum of what I need to do to prep for upgrade tests, but there's plenty of other cleanup needed. |
|
GCE e2e build/test failed for commit c3772563f434ff90f9b24a0f3b6d971edc65977e. Please reference the list of currently known flakes when examining this failure. If you request a re-test, you must reference the issue describing the flake. |
|
GCE e2e build/test passed for commit 9393b6e03ea3d3fa9cb10d7ef192a9ae3e267ea0. |
|
GCE e2e build/test passed for commit 8e1aedb9bb458a4e2a2b3b92bda5458764419147. |
| echo "Found KUBE_RUN_FROM_OUTPUT=y; will use binaries from _output" | ||
| cp _output/release-tars/kubernetes*.tar.gz . | ||
| function check_dirty_workspace() { | ||
| if [[ ! "${JENKINS_TOLERATE_DIRTY_WORKSPACE:-}" =~ ^[yY]$ ]]; then |
|
@spxtr Yup, you're right; fixed. PTAL. |
|
Progress toward #22104. |
|
GCE e2e build/test passed for commit 29506149e32b393694df3d4af585685cbeb33263. |
|
Rebased to pick up fix for #22128. |
|
@spxtr PTAL. |
|
GCE e2e build/test passed for commit c9132e775037c4b92b19bdec139d94f960327d4f. |
|
LGTM, where the G stands for Great! As usual, we might've missed something and this might break some Jenkins stuff. Probably squash and manual merge? @ixdy you might be interested as well. |
|
@spxtr squashed. Let's manually merge tomorrow morning. |
|
SGTM |
|
GCE e2e build/test passed for commit adbd8a9. |
ikehz
added
the
lgtm
|
Applying LGTM per above comments & merging manually to watch Jenkins. cc @k8s-oncall. |
Prep for upgrade job infra fixups
|
@k8s-oncall @ihmccreery heads-up about breakage I observed (although I'm reusing this script in a way it shouldn't be used, so we'll see how Jenkins does) |
Right now, this is just work in preparation for making upgrade jobs easier to run in isolation, instead of having the 7-step mess.
I don't love how I structured
JENKINS_USE_EXISTING_BINARIESandJENKINS_TOLERATE_DIRTY_WORKSPACE, but I couldn't figure out anything better. Feedback encouraged.Plan:
For each
X-Yversion skew, run 4 suites:Xand run Kubectl tests atYX, upgrade master toY, and run tests atXX, upgrade master and nodes toY, and run tests atXX, upgrade master and nodes toY, and run tests atYThis should give us the same coverage as we have now, but without chaining runs together. I need this refactoring because I'll have to fetch and unpack different tests between cluster-up and test.