Describe the bug: ...

A parent span being a DroppedSpan is causing an AttributeError to be called within end_span. I cannot find the reason why. Randomly started happening at 4am on a Sunday (in other words not connected to rollout). Happened across many different task contexts. Believe it was connected to some connection issues with the instrumented code. In anycase the code should protect against this type of problem

Here is the trace from Sentry:

To Reproduce

I have not been able to reproduce, but changing the check to a hasattr(parent, "id") should be sufficient.

Expected behavior: ...

Don't throw errors when ending span.

Environment (please complete the following information)

• OS: [e.g. Linux] Ubuntu
• Python version: 2.7
• Framework and version [e.g. Django 2.1]: Flask 0.12
• APM Server version: 6.6.1
• Agent version: 4.1.0

Additional context

Add any other context about the problem here.

• Agent config options

  Click to expand

  TRANSACTIONS_IGNORE_PATTERNS = ["^OPTIONS "]
  SERVER_URL = "http://apm:8200"

• requirements.txt:

  Click to expand

  elastic-apm==4.1.0
  Flask==0.12