Rename error_handler settings to on_error

[TonyCTHsu]

2.0 Upgrade Guide notes

The error_handler settings have been replaced by on_error to align with the options for Datadog::Tracing.trace(options).

Changing the name in your configuration from error_handler to on_error is all that is required for most integrations.

Complete details are below.

integration	changes
active_job	Remove error_handler, it was never used.
delayed_job	Rename error_handler to on_error
excon	1. Remove error_handler 2. Add on_error and error_status_codes
faraday	1. Remove error_handler 2. Add on_error and error_status_codes
grpc	1. Remove error_handler, server_error_handler and client_error_handler 2. Add on_error.
pg	Rename error_handler to on_error
que	Rename error_handler to on_error
resque	Rename error_handler to on_error
shoryuken	Rename error_handler to on_error
sidekiq	Rename error_handler to on_error
sneakers	Rename error_handler to on_error

For GRPC: on_error would be invoke on both server and client side instrumentation. Merge your server_error_handler and client_error_handler to on_error. The implementation for on_error should distinguish between the server and client.

Datadog.configure do |c|
  c.tracing.instrument :grpc, 
     on_error: proc do |span, error|
        if span.name == 'grpc.service'
           # Do something for server instrumentation
        end
        if span.name == 'grpc.client'
           # Do something for client instrumentation
        end
     end
end

For faraday and excon:

• Remove error_handler settings
• Configure error_status_codes to control tagging span with error, the default is 400...600 (previous only server servers are tagged.)
• Configure on_error to control behaviour when exception is raised. For example: Faraday::ConnectionFailed/Excon::Error::Timeout

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (c5692b8) 98.06% compared to head (807abc2) 98.09%.
Report is 4 commits behind head on 2.0.

Additional details and impacted files

@@            Coverage Diff             @@
##              2.0    #3341      +/-   ##
==========================================
+ Coverage   98.06%   98.09%   +0.03%     
==========================================
  Files        1241     1248       +7     
  Lines       72027    72129     +102     
  Branches     3395     3397       +2     
==========================================
+ Hits        70632    70757     +125     
+ Misses       1395     1372      -23     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ekump]

Upgrade note suggestions:

Rename the error_handler settings to on_error to align with the options for Datadog::Tracing.trace(options) . The on_error handler is invoke when the instrumented method raises an exception.

The error_handler settings have been replaced by on_error to align with the options for Datadog::Tracing.trace(options).

For most integrations, change your configurations options from error_handler to on_error.

I would move this line above the table and change it to:

Changing the handler name in your configuration from error_handler to on_error is all that is required for most integrations. Complete details are below.

For GRPC: on_error would be invoke on both server and client side instrumentation. Merge your server_error_handler and client_error_handler to on_error. The implementation for on_error should distinguish between the span.

For GRPC: on_error will be invoked on both server and client side instrumentation. Merge your server_error_handler and client_error_handler logic to on_error. The implementation for on_error should distinguish between the span.

I'm not exactly sure what "The implementation for on_error should distinguish between the span." means?