[MESOS-8467] Destroyed executors might be used after `Slave::publishResource()`.

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Fix Version/s: 1.7.3, 1.8.0
• Labels: mesosphere mesosphere-dss-ga storage
• Target Version/s: 1.7.3, 1.8.0
• Epic Link: Resource Provider and CSI Tech Debt
• Sprint: Storage R11 Sprint 40, Storage: RI-13 Sprint 44
• Story Points: 2
• Assignee: Chun-Hung Hsiao
• Reporter: Chun-Hung Hsiao
• Shepherd: Meng Zhu
• Created: 19/Jan/18 18:04
• Updated: 10/Apr/19 21:37
• Resolved: 10/Apr/19 21:37

Description

In the following code from https://github.com/apache/mesos/blob/7b30b9ccd63dbcd3375e012dae6e2ffb9dc6a79f/src/slave/slave.cpp#L2652:

publishResources()
  .then(defer(self(), [=] {
    return containerizer->update(
        executor->containerId,
        executor->allocatedResources());
  }))

A destroyed executor might be dereferenced if it has been move to Framework.completedExecutors and kicked out from this circular buffer. We should refactor Slave::publishResources() and its uses to make the code less fragile.

Issue Links

• is related to: MESOS-9711 Avoid shutting down executors registering before a required resource provider.