[MESOS-6785] CHECK failure on duplicate task IDs

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Component/s: master
• Labels: mesosphere
• Story Points: 3
• Assignee: Unassigned
• Reporter: Neil Conway
• Shepherd: Vinod Kone
• Created: 12/Dec/16 22:18
• Updated: 12/Jun/17 23:06
• Resolved: 12/Jun/17 23:06

Description

The master crashes with a CHECK failure in the following scenario:

1. Framework launches task X on agent A1. The framework may or may not be partition-aware; let's assume it is not partition-aware.
2. A1 becomes partitioned from the master.
3. Framework launches task X on agent A2.
4. Master fails over.
5. Agents A1 and A2 both re-register with the master. Because the master has failed over, the task on A1 is not terminated ("non-strict registry semantics").

This results in two running tasks with the same ID, which causes a master CHECK failure among other badness:

master.hpp:2299] Check failed: !tasks.contains(task->task_id()) Duplicate task b88153a2-571a-41e7-9e9b-c297fef4f3cd of framework eaef1879-8cc9-412f-928d-86c9925a7abb-0000

Issue Links

• duplicates: MESOS-3070 Master CHECK failure if a framework uses duplicated task id.
• relates to: MESOS-6805 Check unreachable task cache for task ID collisions on launch