We have an AWS Lambda running in Go, and upon initialisation runs the following to initialise AWS X-Ray
err := xray.Configure(xray.Config{
LogLevel: "info",
ServiceVersion: "1.2.3",
})
In a seperate repository, we have a utils repository which exposes a HTTP library for our internal stuff. This is imported as a git submodule to all other Lambdas. The code is as follows:
ctx, subseg := xray.BeginSubsegment(incomingContext, "Outbound HTTP call")
client := xray.Client(&http.Client{Transport: tr})
// further down
client.Do(req)
// finally
subseg.Close(resp)
This works as expected when deployed on AWS, producing a nice graph.
The problem is running unit tests on the utils repository. In the context of that repository alone, X-Ray has not been configured, so on the BeginSubsegment call I get a panic:
panic: failed to begin subsegment named 'Outbound HTTP call': segment cannot be found.
I want to gracefully handle the case when X-Ray has not been configured, log it, and carry on execution regardless.
How can I ensure to properly error handle the call to BeginSubsegment when it does not return an error object?
In the case of lambda this code executes without any panic is because lambda creates a facade segment and then your code will be creating subsegments. In the non lambda environment you will have to create a segment first before creating a subsegment. If you don't then it will generate a panic. Now, If you want to log this panic and continue executing your unit tests then I would recommend you to set AWS_XRAY_CONTEXT_MISSING environment variable to LOG_ERROR. It will basically log your panic and continue executing your unit tests.