You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is an overarching issue/tracker for the progress on delivering the transparent mgmt datapath
When vrnetlab spawns a container that runs a VM inside it uses qemu user networking mode to "stitch" the container's eth0 (management) interface to a management interface of a VM.
This approach has a major drawback of using a "fake" IP address assigned to the VM's management interface from the IP range of the qemu user network, which is in vrnetlab set to 10.0.0.0/24. Because of this, all VMs spawned by vrnetlab/containerlab have the same IP (10.0.0.15/24) assigned to their management interface.
This, of course, is not ideal, and it a big pita when trying to manage the containerlab lab nodes with some NMS, as they typically use the IP addresses from the mgmt interface to identify the nodes.
Thanks to @vista- work in #268 there has been added a mode to vrnetlab where the interface of a wrapping container (eth0) is transparently mapped (using tc) to the VM't tap0 interface, which allows the VMs to use the same IP address for its management interface as the eth0 interface of the wrapping container. Read more details about this in #268
Since this change is quite major, it is toggle-able, by setting the CLAB_MGMT_PASSTHROUGH env var to a string value of "true" to enable this new behavior.
The main development PR is #287. It should be used as a base for other vendor implementations. When we are happy with the coverage we will merge #287 into master.
Support for vendors
Nokia
SR OS
Nokia SR OS has to be refactored to support this mode, the work is done by @michelredondo in #272
When vrnetlab spawns a container that runs a VM inside it uses qemu user networking mode to "stitch" the container's
eth0
(management) interface to a management interface of a VM.This approach has a major drawback of using a "fake" IP address assigned to the VM's management interface from the IP range of the qemu user network, which is in vrnetlab set to 10.0.0.0/24. Because of this, all VMs spawned by vrnetlab/containerlab have the same IP (10.0.0.15/24) assigned to their management interface.
This, of course, is not ideal, and it a big pita when trying to manage the containerlab lab nodes with some NMS, as they typically use the IP addresses from the mgmt interface to identify the nodes.
Thanks to @vista- work in #268 there has been added a mode to vrnetlab where the interface of a wrapping container (eth0) is transparently mapped (using tc) to the VM't tap0 interface, which allows the VMs to use the same IP address for its management interface as the
eth0
interface of the wrapping container. Read more details about this in #268Since this change is quite major, it is toggle-able, by setting the
CLAB_MGMT_PASSTHROUGH
env var to a string value of"true"
to enable this new behavior.The main development PR is #287. It should be used as a base for other vendor implementations. When we are happy with the coverage we will merge #287 into master.
Support for vendors
Nokia
SR OS
Nokia SR OS has to be refactored to support this mode, the work is done by @michelredondo in #272
Juniper
vJUNOS-X
Initial delivery in #268
Juniper vSRX
Added in #288
Cisco
all of the above added in #290 by @kaelemc
SONiC
Huawei
Arista
MikroTik
Dell
The text was updated successfully, but these errors were encountered: