[MESOS-9883] Check container timeout counting should be started when the check command is executed.

• Type: Improvement
• Status: Accepted
• Priority: Major
• Resolution: Unresolved
• Component/s: containerization
• Labels: containerization foundations mesosphere
• Epic Link: TaskCheckingImprovements
• Assignee: Unassigned
• Reporter: Gilbert Song
• Created: 09/Jul/19 00:02
• Updated: 20/Aug/19 16:32

Description

Right now the check container timeout counting is started when the checker process sends the request to the agent API. This may not be what users are expecting, and sometimes it may take longer for the agent process to launch a check container if the agent is under a heavy workload. This could lead to the health check being failed if the timeout is small and the agent is slow.

In common sense, users may expect the timeout to be just the timeframe that the command should finish. So that users could better define the health check timeout.

Issue Links

• is blocked by: MESOS-9945 Use streaming response in the checker process
• is duplicated by: MESOS-9944 Command health check timeout begins too early