Subnet should allow for disabling DNS nameservers (instead of pushing dhcp ip as default)
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Expired
|
Low
|
Unassigned |
Bug Description
When connecting a VM to more than 1 network interface, defaults of the second subnet will override user-defined settings of the first (usually primary) interface.
Reproduce:
1. create a VM with 2 network interfaces where:
eth0 - subnet with a GW, and a custom DNS nameserver
eth1 - secondary network where subnet is created with default settings, dhcp enabled
NOTE: most images will require manually requesting DHCP on eth1
2. check routing and DNS details on VM.
custom settings from primary subnet's DNS, and default GW have been overridden by secondary subnet's defaults
Workarounds:
1. reverse network settings so that primary interface sends DHCP request last.
- problematic because usually the most important network and NIC should be defined first. Also, some VMs might be connected only to primary, so this would create inconsistencies between VMs in the same network
2. Manually disable defaults on secondary subnet:
- Works for GW.
- Doesn't work for DNS, since Neutron configures dnsmasq to push DHCP port's IP when no DNS nameserver is defined
3. Manually set secondary subnet's DNS to match primary's.
- Not all users have access to the data of all other subnets. Primary network might have been created by another user
Changed in neutron: | |
assignee: | nobody → Satish Inakoti (inakoti-satish) |
Changed in neutron: | |
assignee: | Satish Inakoti (inakoti-satish) → Avinash (avinash-jalumuru) |
Changed in neutron: | |
assignee: | Avinash (avinash-jalumuru) → nobody |
The issues doesn't seem to have an obvious solution due to the fact that two independent subnet settings conflict when being used by one VM.
I'm not sure there is any problem on neutron side.
Are there any suggestions? It looks like the only way to fix this is to introduce proper API support which might be quite complex.