OpenStack's approach to monitoring VM's has traditionally been a very 'black box' approach. VNF Managers have played the role of mapping between a VM's Application-Specific state and OpenStack's management of the VM. Recent work in the OpenStack Masakari project has started to look at leveraging a simple standardized API between an OpenStack compute host and it's hosted Guest VMs. Such work enables capabilities as 'active' / white-box type monitoring of the VM and Applications within the VM, graceful notifications of potentially service-affecting administrative commands on the VM, and even the ability to NACK these potentially service-affecting commands if the HA VM is not fully synchronized with it's peer VM. In this session we'll discuss the pros and cons of such an approach, and provide a simple demo of these capabilities.
Recent work in the OpenStack Masakari project has started to look at leveraging a simple standardized API between an OpenStack compute host and it's hosted Guest VMs. Such work enables capabilities as 'active' / white-box type monitoring of the VM and Applications within the VM, graceful notifications of potentially service-affecting administrative commands on the VM, and even the ability to NACK these potentially service-affecting commands if the HA VM is not fully synchronized with it's peer VM.
