[MESOS-4824] "filesystem/linux" isolator does not unmount orphaned persistent volumes

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 0.24.0, 0.25.0, 0.26.0, 0.27.0
• Fix Version/s: 0.28.0
• Component/s: containerization
• Labels: containerizer mesosphere persistent-volumes
• Epic Link: Persistence
• Sprint: Mesosphere Sprint 30
• Story Points: 2
• Assignee: Joseph Wu
• Reporter: Joseph Wu
• Shepherd: Jie Yu
• Created: 01/Mar/16 00:28
• Updated: 26/Nov/18 12:22
• Resolved: 26/Nov/18 12:22

Description

A persistent volume can be orphaned when:

1. A framework registers with checkpointing enabled.
2. The framework starts a task + a persistent volume.
3. The agent exits. The task continues to run.
4. Something wipes the agent's meta directory. This removes the checkpointed framework info from the agent.
5. The agent comes back and recovers. The framework for the task is not found, so the task is considered orphaned now.

The agent currently does not unmount the persistent volume, saying (with GLOG_v=1)

I0229 23:55:42.078940  5635 linux.cpp:711] Ignoring cleanup request for unknown container: a35189d3-85d5-4d02-b568-67f675b6dc97

Test implemented here: https://reviews.apache.org/r/44122/