gridscale / gridscale_api_client_python Goto Github PK

To make this client more in line with the official API I'd to propose a rename of the following methods:

• get_ssh_keys
• get_ssh_key
• create_ssh_key
• update_ssh_key
• delete_ssh_key
• get_ssh_key_events

ssh_key is canonically without the underscore in the API. The new function names would then be

• get_sshkeys
• get_sshkey
• create_sshkey
• update_sshkey
• delete_sshkey
• get_sshkey_events

For backwards-compat we could alias the methods to their old names.

It's a small annoyance, but I still wanted to mention it. I'm ready to work an a pull request if this change finds acceptance.

We want to create a new config file that is shared across the different libraries and tools that use the gridscale API: gridscale/terraform-provider-gridscale#183

For this we need to have a function that reads that config file into an internal representation and probably also a function that writes such a config file if it does not yet exists.

See gridscale/terraform-provider-gridscale#183 for the format of the file and file paths on the different platforms.

Steps to reproduce:

Get an account with no buckets in object storage. curl the public API and verify:

$ GS_USER_ID=xyz; GS_API_TOKEN=xyz; curl https://api.gridscale.cloud/objects/objectstorages/buckets -H "X-Auth-UserID: $GS_USER_ID" -H "X-Auth-Token: $GS_API_TOKEN"
{"buckets": []}

Run test script (see below):

$ GS_USER_ID=xyz GS_API_TOKEN=xyz python test.py{'buckets': None}

So far so good. Now create a bucket in the panel, let say 'fubar' and curl again.

$ GS_USER_ID=xyz; GS_API_TOKEN=xyz; curl https://api.gridscale.cloud/objects/objectstorages/buckets -H "X-Auth-UserID: $GS_USER_ID" -H "X-Auth-Token: $GS_API_TOKEN"
{"buckets": [{"name": "fubar", "usage": {"num_objects": 0, "size_kb": 0}}]}

Looks good.

Now the API client:

$ GS_USER_ID=xyz GS_API_TOKEN=xyz python test.py
{'buckets': None}

Looks not okay.

If you add more buckets, get_buckets keeps returning none.

Test script (test.py):

import os
from pprint import pprint

from gs_api_client import Configuration, SyncGridscaleApiClient


user_id = os.getenv("GS_USER_ID")
assert user_id

api_token = os.getenv("GS_API_TOKEN")
assert api_token


config = Configuration()
config.host = "https://api.gridscale.cloud"
config.api_key["X-Auth-Token"] = api_token
config.api_key["X-Auth-UserId"] = user_id
config.debug = False

client = SyncGridscaleApiClient(configuration=config)
pprint(client.get_buckets())

Currently the synchronous API client will poll asynchronous requests every 0.5s. This makes hitting the API rate limit quite likely once you introduce concurrency.

For the API client to not feel sluggish, the poll intervall likely shouldn't simply be increased to a few seconds. It could be dynamically increased to a max and/or be request type-dependent.

In the example file it currently tries to use Debian 9 as a template.

We could just change to the newest Debian version for the example. However, this will just break whenever we depreciate Debian 11.

Another option would be to implement a default option which would always use the newest Debian version.

Since the introduction of API rate limiting, requests failed with an ApiException when hitting the limit (429).

I would propose that the client handles retries in that case.

when using the example file it does not work because by default Debian 9 is selected.

We want to switch the default branch name from master to main for all gridscale Open Source projects including this repository.

This involves:

• Switching the default branch name here on GitHub
• Updating all references of the master branch name in GH Actions and other CI/CD pipelines and scripts
• Giving contributors a bit of info on how to update their remotes in their local copies of the repo

Would be nice to do this coordinated and in "one go" so that contributors are not surprised by inconsistent naming or repositories not having switched yet. If you want to help out here contact @bkircher

Regarding to a customer's feature request, I, @bkircher , and the customers really want to have shared-credential feature in all gridscale tools (which are built on top of gridscale public API). The proposal is posted here Proposal of shared credentials across gridscale tools. Please feel free to comment.

Add GitHub Action that executes all Python test cases upon push to branch. This should be pretty simple.

Rough sketch:

• https://stackoverflow.com/q/60298449/104274

Docs:

• https://docs.github.com/en/actions

Hey there. I would like this repository to be renamed to include dashes instead of underlines, i.e. instead of "gridscale_api_client_python" we would name it "gridscale-api-client-python".

Reasons:

• Would be consistent with other repo names in gridscale organization but also on GH in general
• And https://www.youtube.com/watch?v=AQcSFsQyct8

Any objections?

API client should support custom certificates, which will be are used by LBaaS.

Right now there is no config for any linter in this project.

Things to do:

• Choose a linter
• Create a config for that linter
• Make lint run pass without warnings (either by fixing issues in code or by silencing individual warnings in the config)
• Add GH Action task that runs this linter with that config on the project

We could use pylint or flake8. Personally, I would prefer pylint because I know it best but no hard feelings. I could gift an initial config file for pylint.