Skip to content

Use debug bundle upload in premerge CI#1018

Merged
pxLi merged 4 commits into
NVIDIA:mainfrom
pxLi:remove-pvc-debug-bundle
Jun 2, 2026
Merged

Use debug bundle upload in premerge CI#1018
pxLi merged 4 commits into
NVIDIA:mainfrom
pxLi:remove-pvc-debug-bundle

Conversation

@pxLi

@pxLi pxLi commented Jun 2, 2026

Copy link
Copy Markdown
Member

Remove PVC usage, use artifactory to persist build info of failure runs.

Signed-off-by: Peixin Li <pxLi@nyu.edu>
@pxLi pxLi self-assigned this Jun 2, 2026
@pxLi pxLi added the build label Jun 2, 2026
@pxLi

pxLi commented Jun 2, 2026

Copy link
Copy Markdown
Member Author

I will run a validation test w/ test-only commit first

pxLi added 2 commits June 2, 2026 09:42
Signed-off-by: Peixin Li <pxLi@nyu.edu>
Signed-off-by: Peixin Li <pxLi@nyu.edu>
@pxLi

pxLi commented Jun 2, 2026

Copy link
Copy Markdown
Member Author

build

@pxLi pxLi marked this pull request as ready for review June 2, 2026 02:47
@pxLi pxLi requested review from NvTimLiu, YanxuanLiu and yinqingh June 2, 2026 02:48
@pxLi

pxLi commented Jun 2, 2026

Copy link
Copy Markdown
Member Author

build

@greptile-apps

greptile-apps Bot commented Jun 2, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

This PR migrates the premerge CI pipeline away from a Persistent Volume Claim (PVC) for inter-stage workspace sharing, relying instead on Jenkins' stash/unstash mechanism (already in use for the source_tree stash) and Artifactory uploads for persisting failure diagnostics.

  • Removes the PVC credential and workspaceVolume persistentVolumeClaimWorkspaceVolume directives from both the Docker-build and GPU-test Kubernetes agents.
  • Adds unstash "source_tree" at the start of the Premerge Test stage so the checked-out repository is available without the shared PVC workspace.
  • Wraps the test execution in a try/catch that calls common.uploadDebugBundle on failure before rethrowing the exception, providing persistent failure artifacts via Artifactory.

Confidence Score: 4/5

The change is safe to merge; the stash/unstash swap for PVC is straightforward and test.sh does not rely on git history, so the default stash exclusions are harmless.

The migration from PVC to stash/unstash is mechanically correct, and the debug-bundle upload on failure is a useful addition. The one concern is that if uploadDebugBundle throws, it silences the original test failure exception rather than chaining errors, which could obscure root causes during debugging — but this is a non-blocking quality issue.

ci/Jenkinsfile.premerge — specifically the catch block around the debug bundle upload.

Important Files Changed

Filename Overview
ci/Jenkinsfile.premerge Removes PVC credential and workspaceVolume from both pipeline stages; replaces shared-filesystem approach with stash/unstash for the test stage and adds a try/catch to upload a debug bundle on failure.

Sequence Diagram

sequenceDiagram
    participant BD as Build Docker Stage (cpu pod)
    participant JS as Jenkins Stash
    participant GT as Premerge Test Stage (gpu pod)
    participant AF as Artifactory

    BD->>BD: git checkout (source)
    BD->>JS: "stash("source_tree", "**")"
    BD->>BD: "build & push Docker image"
    GT->>JS: unstash("source_tree")
    GT->>GT: resolveIncompatibleDriverIssue()
    GT->>GT: bash ci/test.sh pre-merge
    alt test fails
        GT->>AF: uploadDebugBundle()
        GT->>GT: throw original exception
    end
Loading

Reviews (1): Last reviewed commit: "Revert "TEMP: Force premerge debug bundl..." | Re-trigger Greptile

Comment thread ci/Jenkinsfile.premerge
@pxLi pxLi merged commit 1c37853 into NVIDIA:main Jun 2, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants