Skip to content

Release: dedup-skip + requestPublication retry (#450 follow-ups) - #458

Merged
neuromechanist merged 5 commits into
mainfrom
dev
May 13, 2026
Merged

neuromechanist merged 5 commits into
mainfrom
dev

Conversation

@neuromechanist

Copy link
Copy Markdown
Contributor

Release contents

Operational follow-up after merge

  1. Wait for auto-tag → npm publish (v0.8.13).
  2. Push the workflow template (`.github/dataset-workflows/onboard-openneuro.yml` from this repo) into `nemarDatasets/.github/workflows/onboard-openneuro.yml` — picks up the dedup-skip step.
  3. Verify on a fresh ds ID that the workflow now (a) skips already-imported IDs and (b) waits through BIDS validation rather than exiting at `requestPublication`.

Test plan

nemar-bot and others added 5 commits May 13, 2026 06:35
…456)

parse-ids now pre-filters the matrix to skip any ds###### whose on######
counterpart already exists on nemarDatasets with non-empty contents.
Empty shells (size==0, e.g. the #450 recovery state from run 25769577607)
are still kept so operators can re-import after deleting them.

Motivation: until #448 (proper 'update existing OpenNeuro dataset'
command) lands, re-importing a live dataset clobbers state. Yesterday's
run 25769577607 caught this on ds003800 and ds007315 — both had been
successfully onboarded earlier, the re-attempt failed at git-annex
S3 initremote because the bucket already carried the prior annex-uuid.
The current importer's 'Dataset already exists, continuing...' path
silently barrels through enough state to do damage; the cleaner fix is
to never matrix-spawn for those IDs in the first place.

Uses NEMAR_GITHUB_PAT (the existing workflow secret) because nemarDatasets
dataset repos are private.

Refs #450 #448
… runs (#457)

The bounded poll in step 9 only waits for a workflow_run to REGISTER,
not COMPLETE. For larger datasets we then land at requestPublication
while validation is still running, and the backend returns HTTP 422 +
'BIDS validation is currently running'. The importer was exiting
immediately, leaving the dataset stranded with the data pushed but no
publication request filed.

Retry up to 5 times with 5-minute waits (25 min total budget) on that
specific error. Any other error fails fast. Operator can also recover
manually with 'nemar admin publish request <id>' once validation
finishes — that path is now documented in the failure message too.

Refs #450 (caught in run 25782854836 after the original push-order fix)
@neuromechanist
neuromechanist merged commit 4ebb68e into main May 13, 2026
19 checks passed

This branch was previously deployed

2 inactive deployments
development — 4d423545 Deployed May 13, 2026 by neuromechanist via deploy-dev #232
npm — 4d423545 Deployed May 13, 2026 by neuromechanist via publish #791
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant