Kibana: [APM] Use the outcome field to calculate the transaction error rate chart

[cauemarcondes]

For 7.10 the agent team are working on a new field called outcome (https://github.com/elastic/apm/pull/299) which will be available on the Transaction and Span indices. The field represents a success or a failure transaction or span.

Possible results:

• "failure": Indicates that this transaction describes a failed result
• "success": Indicates that this transaction describes a successful result.
• "unknown": Indicates that there's no information about the outcome. (Won't be counted as error)

Solution
total_number_of_transactions_with_errors / total_number_of_transactions

Fallback:
Should we consider keep the current implementation as a fallback in case of the field is not available?

[elasticmachine]

Pinging @elastic/apm-ui (Team:apm)

[cauemarcondes]

Blocked until the agents have implemented the new field.

[felixbarny]

@beniwohli is working on the reference implementation for elastic/apm#299

@cauemarcondes and @beniwohli please sync up to ensure we have a smooth end-to-end experience.

[felixbarny]

I think this issue is unblocked now that we have specified that we'll use event.outcome to determine whether or not an event is a failure.

While testing against real data is important, I don't think it blocks the UI implementation. You could fake the data by setting the event.outcome on existing documents.

[cauemarcondes]

@felixbarny can you confirm that I'll need to use both Transaction and Span indices to calculate the error rate since both will have event.outcome?

[felixbarny]

No, to calculate the error rate you only have to account for the transaction errors.

The reason is that a transaction can be successful in spite of having an erroneous span. For example, if an API call failed but the server successfully retried or used a fallback.

[cauemarcondes]

No, to calculate the error rate you only have to account for the transaction errors.

Cool, thanks for the explanation.

And about the RUM agent (As there's no browser API to get the status code of a page load, the RUM agent always reports"unknown"for those transactions.), should I hide the error rate on its page?

[felixbarny]

That’s only true for pageload transactions, not Ajax calls. So I wouldn’t hide it. Plus, users may use the API to even mark a pageload transaction as failed.