[MESOS-9482] Resource provider manager can crash on invalid data from resource providers

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.9.0
• Labels: mesosphere mesosphere-dss-post-ga storage
• Epic Link: Resource Provider and CSI Tech Debt
• Sprint: Storage: RI-17 Sprint 52
• Assignee: Benjamin Bannier
• Reporter: Benjamin Bannier
• Shepherd: Chun-Hung Hsiao
• Created: 17/Dec/18 09:53
• Updated: 24/Aug/19 21:24
• Resolved: 24/Aug/19 21:24

Description

The resource provider manager code currently contains a number of assertions which will crash the manager (and its agent) if some forms of invalid data are received from a resource provider. This is dangerous since resource providers are not necessarily part of Mesos-controlled code (they talk to the manager over an HTTP API and could even be in external processes).

Instead of crashing, the resource provider manager should disconnect the resource providers in such scenarios.

Issue Links

• is duplicated by: MESOS-9407 `UPDATE_STATE` with bad resource provider IDs will crash the agent.