[Dashboard] Improve handling of logs and errors in dashboard backend

[mitchellstern]

Why are these changes needed?

This addresses an efficiency issue in the initial version of the dashboard backend wherein the full set of logs and errors were being sent on every update. This PR introduces a better solution where only the log and error counts are sent by default, with two new API routes for retrieving logs and errors themselves (either for a specific (IP address, PID) pair or everything for a single IP address).

This PR moreover unifies the backend handling of logs and errors through the consistent use of IP addresses as identifiers, instead of using IP addresses for logs and hostnames for errors.

In addition, a minor discrepancy on some systems in the reporter is fixed by replacing a the custom determine_ip_address() function with a call to ray.services.get_node_ip_address().

CC @simon-mo @robertnishihara

Related issue number

N/A

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://ray.readthedocs.io/en/latest/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failure rates at https://ray-travis-tracker.herokuapp.com/.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/17533/
Test FAILed.

[mitchellstern]

This PR also includes some minor UI updates.

General dashboard:

Log display:

Error display:

[AmplabJenkins]

Can one of the admins verify this patch?

[simon-mo]

Few minor comments. Otherwise looks good! I really like the massive refactoring of hostname -> ip

[simon-mo]

jenkins add to whitelist

[virtualluke]

Was anything done to remove the token, or is anyone doing anything interesting with say nginx to make access to the webserver easier than reconfiguring nginx each time? I mean I could of course set up a reverse tunnel each time as well but would be better if the uri anchors worked long-term in a reverse proxy (like nginx).

[mitchellstern]

Was anything done to remove the token, or is anyone doing anything interesting with say nginx to make access to the webserver easier than reconfiguring nginx each time? I mean I could of course set up a reverse tunnel each time as well but would be better if the uri anchors worked long-term in a reverse proxy (like nginx).

Hi @virtualluke, we'll be removing the token and addressing some of the other concerns brought up on the Ray Slack group in a separate PR right after this one, thanks.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/17565/
Test FAILed.