[A0CLI-5] feat: logs tail

[shushen]

Description

Add logs tail command

The implementation is naive: it starts with a search query sorted by date:1 and then takes the last logID and switch to checkpoint method in a loop.

The limit of 100 log entries returned in each API call and the sleep duration of 1 second potentially could risk falling behind if the log is being generated at a rate > 100/second, therefore, we only sleep when we are getting less than 90 log entries.

References

https://auth0team.atlassian.net/browse/A0CLI-5

Testing

• This change adds test coverage for new/changed/fixed functionality

Checklist

• I have added documentation for new/changed functionality in this PR or in auth0.com/docs
• All active GitHub checks for tests, formatting, and security are passing
• The correct base branch is being used, if not master

[cyx]

Have a bunch of nits, but otherwise 💪 🙇

[paddycarey]

Looks good! Left some comments and a suggestion for the formatter/renderer.

However the command is broken currently due to the list assignment issue mentioned, that should be fixed before merging.

As an aside, it shouldn't block this PR, we can change it later if wanted, but I quite like how the heroku CLI does logs i.e. not with a tail subcommand, but rather a couple flags on logs itself:

So it'd look like:

• auth0 logs (return 100 recent logs)
• auth0 logs -n $x (return $x recent logs)
• auth0 logs --tail (tail logs)

What do you think?

[paddycarey]

I was thinking about this one, should we just remove the sleep and let the SDK handle retrying and rate limits? It'll make the initial request quickly but gracefully back off if we're rate limited for whatever reason. NAB.

[shushen]

Hitting rate limit itself triggers an event and generates an additional log entry, I'm inclined to avoid hitting rate limit if possible. Since this is not real-time logs, the event itself also only appears after a short interval, there is probably not very useful in trying to be real-time.

I've changed the implementation to only sleep when we are not getting a lot entries (<90) returned, otherwise we will keep polling. This should avoid falling behind as I noted earlier.

[shushen]

As an aside, it shouldn't block this PR, we can change it later if wanted, but I quite like how the heroku CLI does logs i.e. not with a tail subcommand, but rather a couple flags on logs itself:

So it'd look like:

• auth0 logs (return 100 recent logs)
• auth0 logs -n $x (return $x recent logs)
• auth0 logs --tail (tail logs)

I like that! This PR was meant to get things started and was obviously pretty rough. Sorry for stealing the thunder with my mediocre implementation :-) and thanks a lot for the review! Please take another look!

[paddycarey]

LGTM