31/08/23 - Openstack cleanup, dashboards and API work
Migrated all API repos to use ska-src permissions pypi package on gitlab (and added CI to build this package). Used guide at docs.gitlab.com/ee/user/packages/pypi_repository.
Migrated storm-webdav from 'c' flavour: https://confluence.skatelescope.org/display/SRCSC/STFC+Cloud+instance+cleanup
Added dashboard for Coral's aggregated transfer events: https://monit.srcdev.skao.int/grafana/d/XIht_fC4k/coral-rmb-scratch?orgId=1
30/08/23 - API work and machine cleanup
Watched in on "c" flavour machines on IRIS being decommissioned.
API work done:
- Added bin/ directory to repository tree diagrams in READMEs.
- Added scope authz to permissions API (required by service-to-service client_creds flows) which will allow for grafana to access data management endpoints dynamically.
- Login tokens now require correct scopes like we do with audience.
- Permissions module checks that the issuer of the tokens is the SKA IAM.
Started work on publishing permissions api (for client) to gitlab pypi repository via CI. Will need to remove submodules and have this dynamically pull in other dependent repos (at the moment, this clone is housed in the dockerfile, it will need to be in requirements and an --index-url specified. This falls back to PyPi).
29/08/23 - Back up to speed, APIs
Looking through PI20 planning etc.
Data management /locate endpoint now returns site also. Amended API docs to reflect this.
17/08/23 -> 28/08/23 A/L
16/08/23 - APIs and IRIS OpenStack clean up
Checked dmgmt dependencies added back in and works & deployed.
Cleaned up openstack instances.
15/08/23 - APIs
- Added docstring for sited endpoints
- Updated sited README
- Made api doc for site-capabilities-api
- /sites/{site} not working on site-capabilities (multiple version keywords going through permissions client if path_params contains same keywords as check_permission methods). Resolution: not expanding kwargs anymore.
- Updated local/remote git repos following reorganisation of gitlab (sited, datalake api)
- Deployed api services using external domains + redirected ivoa old domains
- Updated referencing for site-capabilities (e.g. ivoa work) in rucio-datalink service
- Fixed datalink helm deployment in wrong namespace.
- Data mgmt API now uses HTTPBearer to manage header authz header
- Amended data mgmt README
- Added group structure for data-management API
- Data mgmt refactored to temporarily use rucio admin access token (token exch not possible with two different iams)
11/08/23 - APIs
- hidden version param (default==latest)
- 401/403 distinction
- datalake api doc refactor
- helping with datalake epic for PI20
- started sited API doc, but encountered problem with multiple version keywords
10/08/23 - APIs
More refactoring:
- permissions api /authorise/exchange now returns the audience for the service from the schema in the response
- added whoami on authn api
- added submoduled permissions client to authn api
- removed token exchange stuff from permissions (done? double check)
- api for authn made + README adjusted
09/08/23 - APIs
More refactoring:
- added cache
- added authn api
08/08/23 - APIs
More refactoring:
- add ping for permissions client + move body logic + version from VERSION file in setup.py
- commit permissions api from site-capababilities submodule (has changes!)
- added token exchange endpoint
- permissions package now manages audience checking + has authorise service endpoint
07/08/23 - APIs
More refactoring:
- Added handle_exceptions
- Permissions API endpoint included and logic to call it instead of verify_permission_for_route()
- Renamed (and group structure)
- Deployment changes: image reg url should include /api, not in template
04/08/23 - A/L
03/08/23 - Sick leave
02/08/23 - Sick leave
01/08/23 - Permissions API
More refactoring:
- Updated permissions README.
- Uploaded permissions API to repo.
- Fixed introspect_token issue (405). To do with a malformed token.
- Made Permissions API doc.
Responding to comments on data management api doc.