Node Access Dependency on Server Manager
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenContrail |
New
|
Medium
|
Abhay Joshi |
Bug Description
This is a system-engineering issue.
I installed Server Manager 2.1 and configured two interfaces on the config and compute nodes, one for management and one for data/control. My intent was to assign static IPs to both interfaces with a fixed IP addresses. The GUI allows the fixed IPs to be configured and the DHCP server subsequently assigns the correct IP addresses to the corresponding interfaces. However the GUI insists on one of the interfaces to be designated as "DHCP". I chose the management interface. After installation on Ubuntu, the /etc/network/
# The primary network interface
auto eth1
iface eth1 inet dhcp
auto eth0
iface eth0 inet static
address 172.16.4.50
netmask 255.255.255.0
~
The issue to consider here is that eth1 is designated as DHCP and when the server manager node fails the config and compute nodes cannot be reached either (1) after a reboot, or (2) after the lease expires. This represents a single point of failure for services with endpoints on the management LAN (e.g., nova). In systems with a single interface server Manager node failure would also affect data traffic.
It is suggested that after initial installation, automatic means be provided to re-configure the dhcp interface as static. This can be implemented as an option in the GUI. As a related feature, the user can be provided with the option of reverting back to the DHCP option for the interface.
tags: | added: server-manager |
Changed in opencontrail: | |
importance: | Undecided → Medium |
assignee: | nobody → Abhay Joshi (abhayj) |