JMX instance configuration might get ignored when there are multiple instances with the same name

[v0lat1le]

We have 2 jmx configurations one in conf.d/jmx.d/conf.yaml:

init_config:
    is_jmx: true
    collect_default_metrics: true
instances:
  - name: riemann
    host: localhost
    port: 7199

and another in conf.d/kafka.d/conf.yaml:

init_config:
    is_jmx: true
    collect_default_metrics: true
instances:
    - name: riemann
      host: localhost
      port: 7199
      refresh_beans_initial: 60

Normally metrics are collected for both of those instances (it is a single application), but sometimes after a deployment (application restart) we completely lose metrics for one of those instances; only restart of the datadog agent resolves the issue.

Using different names for the instances does seem to resolve the issue for us.

I suspect this might be because the instances that fail to fetch metrics are stored in a map keyed on instance name so during an application restart one of the instances would overwrite the other.