Add exclude_last_activity_patterns to prevent some requests from keeping server alive

[minrk]

allows for excluding certain endpoints from activity that might keep the server alive even when it's really idle.

and allow update_last_activity to be a callable

Checklist

• I am the author of this work

What does this PR do?

Type of change:

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation
• Maintenance (testing, packaging, infrastructure, metadata)
• Other

Is this PR related to an issue, or is it part of a larger body of work?

Does this PR introduce a breaking change?

How can this PR be tested?

set exclude_last_activity_patterns and submit requests. last_activity should not be updated.

What should a reviewer concentrate their feedback on?

if this is adequate configuration, or if it should be expressed differently

Other information

cc @shaneknapp who's dealing with lots of idle rstudio sessions not being culled due to constant POST /rstudio/events/get_events

I'll do another PR against rsession-proxy to ignore get_events by default

[shaneknapp]

this is great @minrk! thanks so much!

[ryanlovett]

This will be a useful feature, thanks! A couple comments:

1. In order for downstream proxies to be able to set exclude_last_activity_patterns in their configs (like in the draft PR to jupyter-rsession-proxy), it needs to be a ServerProcess traitlet. I think it should be added to config.py as a traitlet and then passed through in get_proxy_attributes.
2. Should it be documented that when one sets a callable for update_last_activity, it will cause exclude_last_activity_patterns to be ignored?
3. Maybe tests can be added in a later PR.

[minrk]

Added tests and fixed propagation from config, including docstrings explaining that callable update_last_activity is the only thing considered, if defined.

[ryanlovett]

Thanks, looks good. The re.compile in handlers.py can probably be removed too, although I think its a no-op if they're already compiled?

[minrk]

yeah, should be redundant if we can assume the list always comes via validated config. It is definitely a no-op, re.compile(re.compile(re.compile('pattern'))) is fine.

[minrk]

tests seem to have become flaky, some due to waiting for proxy processes to start on Windows, and some of the unix socket tests occasionally failing with ECONNREFUSED. Not sure what's going on there

[minrk]

ok, tests passing after a re-run