Have test coverage for credentials setting
- fix assert
- fix typing
- add a warning on credentials
Sun, Jan 17
- remove useless code as per feeback
- improve tests
Fri, Jan 15
In short, let's KISS!
It would be great to parametrize the request and error type, but unfortunately it does not seem easy to do...
did not see the requested change on clustering keys by vlorentz
TGTM, but needs a look by someone more knowledgeable about the DBs
Use requests.Response.raise_for_status to handle errors generically
Thu, Jan 14
Thanks for the remarks, will go through them.
Wed, Jan 13
Fix copyrights and rebase
refactor auth and backoff
Tue, Jan 12
Mon, Jan 11
Fri, Jan 8
- Take into account feedback from ardumont and vlorentz
- Take into account feedback from douardda
- still TODO: rename "singletons"
Thu, Jan 7
@douardda: taking into account your suggestions.
Wed, Jan 6
- use directory_ls rather than dir_iterator to lower overhead and control iteration
- use a set to test early for known objects
- only update once master head with the given revision id
- do not chunk files ids because in the end the storage requests are not chunked
- partition new and old files to adapt processing
Mon, Jan 4
Wed, Dec 23
Tue, Dec 22
Be more consistent in naming and abbreviating
Take into account feedback from ardumont and vlorentz
Used a "note" directive
Mon, Dec 21
But for components, we wanted to instantiate every instance only one time
What is the benefit of doing that (vs. the cost of the extra complexity)?
First, thanks for the feedback, every help is welcome because I am still pretty the newbie in the project as a whole, and this is a more complex topic that initially anticipated.
Until now, configuration handling was implemented independently for each service.
What do you mean by that? It has always been handled by SWHConfig in swh.core.
btw, you don't have to expose SWHID_RE_RAW directly, it's available as SWHID_RE.pattern
Dec 18 2020
Dec 16 2020
Dec 15 2020
Finally fixed this PREsistent typo!
Reflow the whole source to 88 chars lines
Added a TODO about a section on guidelines
Cleaned and wrote up several sections
Dec 14 2020
In addition to failed cooking retry, failed RPC calls are now also retried.
Dec 11 2020
Removed usage of external host volumes, as asked by @douardda, for brievety.
Keep postgresql without major version pin, as we follow debian versions.
Dec 10 2020
Usa a docker-compose.override definition.
Rephrased where appropriate.
Dec 9 2020
Another strategy preferred: using retrying proxies for storage (existing and activated) and scheduler (to be done).
Dec 8 2020
Tested in docker env, the config is valid after D4685 applied.