Continue on error by default

[liamawhite]

Erroring when able to continue is brittle and was causing the following HELP mismatch on the same metric name to break metric collection merge for us.

# HELP process_start_time_seconds Start time of the process since unix epoch.

# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.

When trying to get merged result from the exporter:

An error has occurred while serving metrics:

collected metric process_start_time_seconds label:<name:"cluster" value:"tcc-staging" > counter:<value:1.58467339842e+09 >  has help "Start time of the process since unix epoch." but should have "Start time of the process since unix epoch in seconds."

Signed-off-by: Liam White liam@tetrate.io

[odeke-em]

Hey @liamawhite, thank you for the PR but I'd like to know how that error came about?

[liamawhite]

This came from using the Prometheus exporter in the Open Telemetry Collector. These were metrics from two different receivers: a prometheus scraper and an opencensus agent.

Those metrics are produced by the respective JVMs from an instance of SkyWalking and an instance of Zipkin. The metrics have the same name but different help fields which causes an error in the Prometheus library, which I presume isn't sure how to merge them.

[odeke-em]

Gotcha, thank you @liamawhite for the explanation and for the PR! LGTM and I'll make a release shortly.