Over provision of stateless PU in case of deploy & destroy timeout resulting in extra instance

Description

under specific circumstances when deploying a stateless PU with a number-of-instances="1" sla the GSM ends up with 2 instances and the deployment status is Intact. The expectation is that the manager tries to eventually match the count of the planned with the running instances. A chronology of what needs to happen 1) The PUI instantiation takes longer than 30 seconds 2) After 30, 60, 90, ... seconds a monitor request is sent to check if instantiation is still running (RequestResponseTimeoutObserver) 3) The monitor request fails due to a timeout 4) GSM tries to remedy this issue by sending a destroy to the PUI 5) destroy times out (lrmi timeout) 6) GSM attempts to allocate the PUI on another GSC 7) PUI finishes instantiation on the first GSC The deployment state is Intact. Planned and Running Instances are shown as 1. 8) PUI finishes instantiation on the second GSC The deployment state is Intact. Planned Instances:1 Running Instances are 2. This state is then kept presumably forever. Reproduction attached.

Attachments

2

Activity

Show:

Details

Assignee

Reporter

Labels

Participants of an issue

Ester Atsmon

Priority

Product

Edition

Platform

All

Affects versions

Freshdesk Support

Open Freshdesk Support
Created August 31, 2022 at 10:06 AM
Updated August 31, 2022 at 10:07 AM
Freshdesk Support