[DigitalOcean] droplet integration

[asaiacai]

This adds digital ocean droplets to the sky.

Tested (run the relevant ones):

• Code formatting: bash format.sh
• All smoke tests: pytest tests/test_smoke.py --do

[moinnadeem]

@asaiacai I would be really interested in using this work -- do you know when you might be able to land this PR?

[asaiacai]

Hello! I’m aiming for this to land in next two weeks providing testing goes smoothly. Stay tuned!

[asaiacai]

@Michaelvll i think this PR is finally ready, once you have a free moment! I've currently disabled tests for gpu droplets as they are still in early access, but I think I got the remaining tests to work on normal CPU droplets fine. Let me know if you find any issues.

[asaiacai]

also this catalog update is required to pass the sky serve requests since I added cheaper 2vcpu instances.

skypilot-org/skypilot-catalog#83

[cblmemo]

Thanks for this amazing work @asaiacai ! This would be very helpful. Left some discussions ;)

[asaiacai]

@cblmemo smoke tests are passing on my end. the catalog PR should fix the CI failures.

[cblmemo]

@cblmemo smoke tests are passing on my end. the catalog PR should fix the CI failures.

Thanks! QQ: Which catalog PR should I merge? Both skypilot-org/skypilot-catalog#83 and skypilot-org/skypilot-catalog#103?

Also, just want to make sure, does the CI passed in your side when merging the above mentioned catalog PRs?

[asaiacai]

only skypilot-org/skypilot-catalog#103

pushed an update to the test_optimizer_dryrun since there are no T4 or 32 memory instances on DO.
rest of the CI passes for me with this change

[cblmemo]

only skypilot-org/skypilot-catalog#103

pushed an update to the test_optimizer_dryrun since there are no T4 or 32 memory instances on DO. rest of the CI passes for me with this change

Just curious, why do has no 32 memory instance? is it because all of them have higher memory?

Also, I think there is other cloud that has no T4 instance. Could you help check why the original test works for them but not do?

$ sky show-gpus T4 --cloud fluidstack
Resources 'T4' not found on Fluidstack. To show available accelerators, run: sky show-gpus --all

[asaiacai]

Just curious, why do has no 32 memory instance? is it because all of them have higher memory?

sorry for the back and forth here. The catalog was missing some instances with 32GB capability, which I just added to the catalog PR only skypilot-org/skypilot-catalog#103 I reverted tests/test_optimizer_dryruns.py back. so the last of the tests here are passing.

Also, I think there is other cloud that has no T4 instance. Could you help check why the original test works for them but not do?
The filter against DO for the T4 case was removed, and the test case passes like Fluidstack now.

[cblmemo]

Thanks @asaiacai ! Just merged the catalog PR. Re-running unittest now and after it passed it should be ready to go!