[UIMA-5814] UIMA-DUCC: agent should support quiesce request

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0-Ducc
• Component/s: DUCC
• Assignee: Jaroslaw Cwiklik
• Reporter: Jaroslaw Cwiklik
• Created: 12/Jul/18 13:45
• Updated: 18/Sep/18 00:22
• Resolved: 18/Sep/18 00:22

Description

Currently when an agent receives a stop request through its admin channel it sends SIGTERM to all its running processes and waits 1 minute for them to stop. Any process still standing after 1 minute is than killed via SIGKILL. 

In addition to stop request, an agent should support a quiesce request (through an admin channel). When such request is received, an agent should send SIGTERM to all its running processes but should not send SIGKILL to allow each process to finish its tasks and exit when done. An agent should not start a timer after sending the SIGTERM.

In quiesce mode, an agent will reject new process allocation requests from the OR. Such process state will be marked as REJECTED and published to the OR. An agent will keep publishing metrics and inventory. 

In an event an agent process does not exit, a sys admin can send a stop request to force termination.

In addition to changes in an agent, modifications are also needed in the RM, perhaps OR, CLI and scripts.