Opentelemetry-go: Fix Flaky CircleCI Builds

[MrAlias]

The Circle CI build system randomly fails to build packages. For example:

• https://circleci.com/gh/open-telemetry/opentelemetry-go/2145
• https://circleci.com/gh/open-telemetry/opentelemetry-go/2133
• https://circleci.com/gh/open-telemetry/opentelemetry-go/2130

The package that fails to build is not consistent, meaning it is not always the same package that fails. An example failure message:

...
FAIL    go.opentelemetry.io/otel/sdk/metric/aggregator/lastvalue [build failed]
...

I have not been able to reproduce this locally, but it definitely needs to be tracked down.

[evantorrie]

Looks like the Go linker is being randomly killed due to OOM. Since OOM killer is inherently a "random" selector, it would explain why this is flaky. Is it possible to upgrade to a bigger docker container to see if that does indeed "fix" the issue?

[MrAlias]

Looks like the Go linker is being randomly killed due to OOM. Since OOM killer is inherently a "random" selector, it would explain why this is flaky. Is it possible to upgrade to a bigger docker container to see if that does indeed "fix" the issue?

@lizthegrey would you know if upgrading our testing instance size is okay? Seems like it is a simple addition of the resource_class property?