[MESOS-8876] Normal exit of Docker container using rexray volume results in TASK_FAILED.

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.4.2, 1.5.1, 1.6.0
• Component/s: docker
• Sprint: Mesosphere Sprint 79
• Story Points: 2
• Assignee: Qian Zhang
• Reporter: Qian Zhang
• Shepherd: Vinod Kone
• Created: 03/May/18 04:20
• Updated: 22/Mar/19 16:41
• Resolved: 07/May/18 21:19

Description

In the fix to  MESOS-8488, we reap the Docker container process directly in Docker executor, and it will wait for `docker run` to return for at most 3 seconds. However, in some cases, the `docker run` command will indeed need more than 3 seconds to return, e.g., the Docker container uses an external rexray volume (see the attached task json as an example), for such container, there will be about 5 seconds between container process exits and the `docker run` returns (I suspect Docker daemon was doing some stuff related to rexray volume during this time), so we will reap this container, and send a TASK_FAILED.

Issue Links

• is caused by: MESOS-8488 Docker bug can cause unkillable tasks.