Expose the API key/host/secret as variables for the users

[wvangeit]

Allows users to use
OSPARC_API_HOST / OSPARC_API_KEY / OSPARC_API_SECRET
in notebooks to access the API.

[wvangeit]

FYI @pcrespov

Related to ITISFoundation/osparc-simcore#5695 (review)

[pcrespov]

thx

[GitHK]

Please make sure OSPARC_API_HOST si resolvable on all deployments before this change goes in

[GitHK]

This does not exist in Osparc codebase ad it will not resolve on all deployments

[wvangeit]

Do you mean this is because the API server is not available on some deployments? Or just because the variable is not set yet?

@pcrespov in your new PR on simcore, what will happen if there is no API server is available? Will this var be empty/None or so, or will it just not be set?

[pcrespov]

@GitHK Actually, waiting for you review ;-) ITISFoundation/osparc-simcore#5695

[pcrespov]

@wvangeit i will then pull it and use it locally.

[pcrespov]

what will happen if there is no API server is available? Will this var be empty/None or so, or will it just not be set?

@wvangeit Good question. In this case, I can see in the code that it simply does not get replaced.

https://github.com/itisfoundation/osparc-simcore/blob/5035fe5cc7a7d7ee9fcab15fd7f8a116a4934cd6/services/director-v2/src/simcore_service_director_v2/modules/dynamic_sidecar/docker_compose_specs.py#L340-L349

Apparently the services will fail upon start if OSPARC_VARIABLE_API_HOST is not exposed in osparc. IMO it is a bit extreme since a normal user (i.e. somebody that is not the service owner) will not understand why this service failed to start. If this expression is an an .env file and OSPARC_VARIABLE_API_HOST is undefined it then OSPARC_API_HOST gets assigned an empty string! Nonetheless @GitHK, hwo is the code owner, has a different view. He will write it down later for you to be aware of it

[GitHK]

Missed your PR. Got distracted. It's all good now.

[wvangeit]

I would vote for empty string. Then the user and the python api client can catch this, and notice that the api host is undefined. (in a sense some 'null' instead of empty string would be better, but not sure if there is a stable null equivalent in the world of env vars)

[GitHK]

I prefer that OSPARC_VARIABLE_* have to be resolvable in order for the service to start. Unlike the OSPARC_VARIABLE_VENDOR_SECRET_*, which are define din the db, these are provided always by the platform.

If you are requesting an OSPARC_VARIABLE_*, which does not exist, this is a clear way to signal that you did something wrong.

Also I think it's the user's responsibility to make sure that the service stars once they make changes to it. He can always test it in a local deployment before publishing it elsewhere. Unfortunately we don't have a better workflow for this part.

[pcrespov]

• FYI vendor secrets are also provided by the service owner.
• directorv2 also fails to start service when the vendor secrets is not available.

[GitHK]

Missed your PR. Got distracted. It's all good now.

[elisabettai]

Very useful feature in the related PR, thanks @pcrespov. If I get this right, if the services specs are adapted correctly, users will not need anymore to create API keys and secrets when they run from inside oSPARC!

I am not sure what the conclusion about OSPARC_API_HOST. I think the obvious thing to avoid is that the service fails to start if this is not set. 😉

Small question: ${OSPARC_VARIABLE_API_HOST} and ${OSPARC_VARIABLE_API_KEY} will always be available, right?

[pcrespov]

@elisabettai yes, they are.
SEE also ITISFoundation/osparc-simcore-clients#149

@elisabettai can you please merge these changes and produce a new version of this service and inject in master registry so we can test it there?. thx

[elisabettai]

To (auto-magically) publish a new version of this: @wvangeit, can you please run make version-patch? Then I'll merge and take care of publishing.

[wvangeit]

@elisabettai could you also deploy it to osparc.io now.
Metamodelling is working there now, I need this variable accessible in the notebooks to test the axondeepseg metamodelling study. Thx.

[wvangeit]

@elisabettai , I ran the version-patch. Thx

[elisabettai]

@pcrespov, @wvangeit, the new version (3.0.2) of JupyterLab Math is now available on osparc-master (and s4l-master).

Let me know if it works as expected and if you need it on other deployments (e.g. osparc.io). jfyi, we haven't published the 3.x.x on osparc.io yet, since we wanted to test the inactivity monitoring in master before publishing it to production.