You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
### Description
- If configured, uploaded datasets are forwarded to S3 after unzipping
- layersToLink now reference the source dataset by datasetId (this
introduces api version 11)
- All newly uploaded datasets are virtual (meaning that the on-disk
datasource-properties.json is removed and the database entry is the
source of truth)
- Only exception: datasets with needsConversion=true. Will be cleaned up
in #8959
- layersToLink no longer creates symlinks but rather just entries in the
DB, just like it works for uploadToPaths
- Raw uploaded data is backed up inside of the .trash folder for
debugging purposes. It gets cleared regularly anyway.
### Steps to test:
- Configure S3 uploads using appropriate credentials in the
application.conf
- Upload a dataset
- View that dataset
- Also test that local filesystem case did not break
- Test uploading zipped dataset and non-zipped dataset
- Test tif upload with `yarn enable-jobs` and a worker
- test upload from libs (ideally with layersToLink)
### TODOs:
- [x] Handle uploading larger files with multipart upload
- [x] Fix uploaded files having "aws-chunked" encoding
- [x] Push updated application.conf without credentials
- [x] Test with more datasets
- [x] Do not upload files that are not referenced in layers (e.g.
datasource properties.json files)
- [x] use UPath for everything
- [x] is there still a case where uploaded datasets should not be
virtual? → no!
- [x] Make all uploaded datasets virtual
- [x] unify layers to link mechanism
- [x] api versioning (layersToLink adapter)
- [x] re-test with s3
- [x] delete local stuff or move to .trash?
- [x] make sure both old and new convert_to_wkw variants are supported
### Issues:
- fixes#8415
- fixes#8934
- fixes#8893
- fixes#8749
- Follow-up issue #8979
------
- [x] Added changelog entry (create a `$PR_NUMBER.md` file in
`unreleased_changes` or use `./tools/create-changelog-entry.py`)
- [x] Added migration guide entry if applicable (edit the same file as
for the changelog)
- [x] Removed dev-only changes
- [x] Considered [common edge
cases](../blob/master/.github/common_edge_cases.md)
- [x] Needs datastore update after deployment
---------
Co-authored-by: Norman Rzepka <[email protected]>
Co-authored-by: Florian M <[email protected]>
Co-authored-by: Florian M <[email protected]>
Co-authored-by: MichaelBuessemeyer <[email protected]>
@@ -654,6 +658,12 @@ class DatasetController @Inject()(userService: UserService,
654
658
dataset.status ==DataSourceStatus.notYetUploadedToPaths || dataset.status ==DataSourceStatus.notYetUploaded) ?~>s"Dataset is not in uploading-to-paths status, got ${dataset.status}."
655
659
_ <-Fox.fromBool(!dataset.isUsable) ?~>s"Dataset is already marked as usable."
0 commit comments