Skip to content

[GRDM-57690] Update user Quota value when moving/copying files - #738

Open
ndnhat1 wants to merge 51 commits into
developfrom
feature/nii_grdm_202603_step4/2.2_quota_calculation_when_move_file
Open

ndnhat1 wants to merge 51 commits into
developfrom
feature/nii_grdm_202603_step4/2.2_quota_calculation_when_move_file

Conversation

@ndnhat1

@ndnhat1 ndnhat1 commented May 7, 2026 •

Copy link
Copy Markdown

Purpose

  • Modify the quota calculation logic when moving/copying files from Extended Storage to Institutional Storage, as well as when moving/copying files between folders within the same storage.
  • Prevent moving or copying files that exceed the maximum storage limit file size.

Changes

  • Update FileInfo for moved/copied files and recalculate the user's quota accordingly.
  • Add a validation step to ensure files do not exceed the max_file_size.

QA Notes

Documentation

Side Effects

Deployment Notes

Please ensure the following branches are deployed to their respective servers:

Ticket

GRDM-57690

@ndnhat1 ndnhat1 changed the title [GRDM-57690 ] Update user Quota value when moving/copying files [GRDM-57690] Update user Quota value when moving/copying files May 7, 2026
tma-hcphat and others added 18 commits May 8, 2026 13:45
…i_grdm_202603_step4/2.2_quota_calculation_when_move_file
 + Update email template handle move/copy file quota and max size error
 + Add script migrate missing file info and update quota used
+ Fix institution default max quota issues (owner lookup, zero quota, quota retrieval, transaction handling)
+ Refactor quota handling logic
+ Update migration (use OneToOneField and remove db_index)
+ Update default value in "Institutional Storage Quota Control" screen
 + Conflict resolve
 + Mail template render
 + Issue data.size is string type
+ Fix user quota update rollback behavior
+ Align related_name between model and migration
+ Remove unused user parameter from file_modified
+ Update UT after change logic
 - Upsert UserQuota on storage type change and skip institution quota for NII storage region
 - Handle NII quota type change on storage credential save and UT

@yacchin1205 yacchin1205 left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

GRDM-57690にコメントを入れました。

@yacchin1205 yacchin1205 left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

OKかと思います。一点、対応するRDM-waterbutlerへのリンク(これを必ずセットでデプロイすること)を、明記いただけると良いかと思います。

@hidefumi-moritani

Copy link
Copy Markdown

OKかと思います。一点、対応するRDM-waterbutlerへのリンク(これを必ずセットでデプロイすること)を、明記いただけると良いかと思います。

@yacchin1205

上記について、明記いたしました。

tishin-endou and others added 12 commits October 2, 2026 15:26
Replace the boto 2 client with boto3/botocore so that the S3 addon can talk
to regions that only accept Signature Version 4 (ap-northeast-1 and later).

- utils.py / views.py / models.py build the client through boto3 with a
  region-specific endpoint, and surface botocore exceptions as HTTPError
- folder_id becomes "bucket:/prefix/" so that a subfolder can be selected;
  the legacy "bucket" form is still accepted
- the node/user config JS, the Rubeus config and the log action lists follow
  the new folder shape
- requirements.txt gains boto3/botocore; addons/s3/requirements.txt is kept
  as an empty file because the Jenkins Dockerfile still does
  COPY ./addons/s3/requirements.txt and would fail the image build without it
- MAX_UPLOAD_SIZE stays at 50GB: multipart upload handles up to 5TB, so the
  lower limit the first draft introduced was unnecessary
- the ${ _("...") } gettext wrappers are kept in all three mako templates
  (20 translatable strings); only the aria-label/alt attributes and the th
  class change are new

Co-Authored-By: An Qiuyu <qiuyu.an@hotmail.com>
…lready registered

Registering the same access key twice has to end up on the existing
ExternalAccount instead of creating a second row. The duplicate is caught by
the unique constraint on (provider, provider_id) — provider_id holds
"{aws_account_id}\t{access_key}" (GRDM-53044) — so the save raises
IntegrityError, not the ValidationError the handler was catching, and the
failed INSERT aborted the surrounding transaction.

Wrap the save in transaction.atomic() so the failure only rolls back to a
savepoint, and catch IntegrityError alongside ValidationError before falling
back to the account that is already stored.
…igV4 folder shape

- test_view.py still used assert_equals/assert_in and a missing
  InstitutionFactory import left over from the nose era; replace them with
  plain pytest asserts against rv.text
- _mock_folder_result still returned the pre-SigV4 get_folders payload, so the
  node settings tests asserted against a shape the view no longer produces
Extend addons/s3/tests/test_view.py so the SigV4 flow is exercised end to end:

- both folder_id forms ("bucket" and "bucket:/prefix/") and a non-existent
  bucket raising InvalidFolderError
- the provider_id format "{aws_account_id}\t{access_key}" and the dedup path
  when the same key is registered twice (GRDM-53044)
- the error branches: a missing key field, get_user_info returning None, an
  invalid bucket name, an empty bucket_exists result, can_list returning None,
  and the ClientError / InvalidAuthError conversions

Three test-harness fixes come with it:

- query ExternalAccount by provider_id, a plain CharField. oauth_key is an
  EncryptedTextField (JWE), so its ciphertext differs on every save and
  filtering on it never matches; the plaintext is checked by attribute access
  instead
- use post_json/put_json instead of the json= keyword, which the webtest_plus
  version used in CI does not accept, and pass expect_errors=True wherever a
  4xx/401/403 response is expected
- patch get_user_info with mock.patch(..., return_value=mock.Mock(...)); the
  .return_value attribute on the patcher object never reaches the started mock,
  so user_info.id / user_info.display_name were missing in the view
…display name

- stop lower-casing the bucket name in the settings check. S3 bucket names are
  case-sensitive in the legacy us-east-1 naming rules, so folding the case can
  point the addon at a different bucket; validate_bucket_name already rejects
  the names that are actually invalid
- fall back gracefully when the owner returned by GetUserInfo has no display
  name, instead of raising while building the account label
- keep the Owner entry of the admin storage-location response in its original
  shape so that the existing consumers keep working
…rametrize

pytest.mark.parametrize has no effect on a method of a unittest.TestCase: the
case is collected once and the extra arguments are never supplied, so
test_serialize_settings_supports_legacy_and_prefixed_folder_ids errored on a
missing fixture instead of checking the three folder_id forms. Iterate over
the cases inside the test body instead.
get_bucket_location() returns a null LocationConstraint for us-east-1,
but BUCKET_LOCATIONS keys that region as '', so the null fell through to
set_folder and labelled the bucket '(None)' in the node settings UI.
The ListObjects call was unguarded, so expired or revoked keys and a
missing bucket escaped as a raw botocore exception and reached the user
as a 500 with no actionable message.  Map NoCredentialsError to 403 and
reuse get_status_for_error so a ClientError keeps the status S3 returned.
create_bucket only handled NoCredentialsError and ClientError, so the
non-response failures - bad parameters, an unreachable endpoint - left
the user with a 500.  ClientError is not a BotoCoreError subclass and
NoCredentialsError is matched earlier, so the existing clauses keep
their behaviour.
Register R2026b at the head of MATLAB_RELEASES and in
MATLAB_PRODUCTNAMES_MAP, and add its mpm input file. It makes the
release available in the BinderHub addon.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
mathworks-ref-arch/matlab-dockerfile@f419020 added a commented-out
supportPackageDestinationFolder section to these files. No product
lines changed, so the generated product name lists are unaffected.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@ndnhat1
ndnhat1 marked this pull request as ready for review October 19, 2026 02:37
…_initial_quota_setting

Feature/nii grdm 202603 step9/2.3 initial quota setting
…api-permission

Feature/fix wiki import api permission
…3-#63370

Feature/binderhub 20261013 #63370
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.

9 participants