Juniper Junos OS Evolved¶
The juniper_evolved model runs Junos OS Evolved as a virtual machine (vJunosEvolved) or as a
container (cJunosEvolved). A node is the virtual machine unless it sets kind = "container".
| Kind | Image | Manifest |
|---|---|---|
virtual_machine |
vJunosEvolved | model = "juniper_evolved" (the default kind) |
container |
cJunosEvolved | model = "juniper_evolved", kind = "container" |
Virtual machine (vJunosEvolved)¶
Base Image¶
Base images can be downloaded from the download page.
Import Image¶
sherpa server image import \
--src /some/path/<image_name>.qcow2 \
--version 25.2R1.8 \
--model juniper_evolved \
--default
Default Node Parameters¶
| Property | Value |
|---|---|
| Kind | Virtual Machine |
| Tested Version | 25.2R1.8 |
| CPU | 4 |
| RAM | 8GB |
| CDROM Driver | SATA |
| DISK | virtioa.qcow2 |
| DISK Driver | virtio |
| ZTP | tftp |
| Management Interface | re0:mgmt-0 |
| Data Interfaces | et-0/0/0-11 |
| Interface Driver | virtio |
Example Manifest¶
name = "juniper-evolved"
[[nodes]]
name = "dev01"
model = "juniper_evolved"
ZTP¶
TFTP¶
The Juniper documentation states that a secondary attached disk is the method to provide an initial
config on first boot. However, I could not get this to work with the latest version: 25.2R1.8.
Therefore, the TFTP method is selected for this device type as default.
Node Access¶
sherpa ssh <node-name>
sherpa console <node-name>
Notes¶
- Approximate time for the system to be ready:
9 minutes
Container (cJunosEvolved)¶
cJunosEvolved runs Junos OS Evolved in a virtual machine inside the container, with a software simulation of a PTX router's forwarding chipset.
Base Image¶
The image can be downloaded from the
vJunos Labs download page. The download is
a docker save archive, for example cJunosEvolved-26.2R1.7-EVO.tar.gz.
Import Image¶
Load the archive into Docker, tag the image with the cJunosEvolved version, and register it with Sherpa by scanning the local container images:
docker load -i cJunosEvolved-26.2R1.7-EVO.tar.gz
docker tag cjunosevolved:26.2R1.7-EVO localrepo/juniper_evolved:26.2R1.7
sherpa server image scan --kind container
Warning
cJunosEvolved images must be tagged with the repo localrepo/juniper_evolved. The scan only
registers images with that repo as juniper_evolved container images, and labs start
cJunosEvolved from it. The first version registered becomes the default.
Every other image command takes --kind container for a cJunosEvolved image; without it, the
command means the vJunosEvolved image. For example:
sherpa server image set-default --model juniper_evolved --kind container --version 26.2R1.7
Platforms¶
cJunosEvolved emulates one of two PTX routers. Choose one per node with
container.platform:
| Platform | Router | Chipset | Data Interfaces |
|---|---|---|---|
bt |
PTX10001-36MR | Express 4 | et-0/0/0-11 |
bx |
PTX10002-36QDD | Express 5 | et-0/0/0-35 |
Without container.platform, a node is bt. A node has 12 data interfaces on either platform unless the
manifest sets interfaces.data_count; a bx node takes up to 36. container.platform applies
to a juniper_evolved container only.
Default Node Parameters¶
| Property | Value |
|---|---|
| Kind | Container |
| Tested Version | 26.2R1.7 |
| CPU | 4 |
| RAM | 6292MB |
| ZTP | volume |
| Management Interface | re0:mgmt-0 |
| Data Interfaces | et-0/0/0-11 |
| Interface Driver | veth |
cpu.count and memory size the Junos OS Evolved virtual machine inside the container. 4 vCPUs
and 6292MB are its minimums. The container needs about 2GB more memory than the virtual machine,
so Sherpa gives it a Docker memory limit of memory plus 2048MB (8340MB by default), and
counts that against the server's memory. It has no CPU quota, so the virtual machine is not
throttled. memory can be at most 63487MB.
Example Manifest¶
name = "juniper-evolved"
[[nodes]]
name = "pe01"
model = "juniper_evolved"
kind = "container"
[[nodes]]
name = "pe02"
model = "juniper_evolved"
kind = "container"
[nodes.container]
platform = "bx"
[[links]]
src = "pe01::et-0/0/0"
dst = "pe02::et-0/0/0"
ZTP¶
Volume¶
Sherpa writes a juniper.conf startup config and mounts its directory at
/home/evo/configdisk in the container. The container hands the config to Junos on first boot.
Requirements¶
- The Sherpa server must have KVM: the container starts a virtual machine. The container runs privileged.
- On first boot the container installs Junos to a virtual disk, so the first start of a node takes longer than later ones.
Node Access¶
sherpa ssh <node-name>
Notes¶
- Junos shows every data interface up, linked or not: its simulated forwarding plane does not follow the link state of the container interfaces. Unlinked interfaces carry no traffic.
- A node shows all of its platform's data interfaces whatever
interfaces.data_countis: 12 onbt, 36 onbx. Only the firstinterfaces.data_countof them are wired to the lab. - Approximate time for the system to be ready:
4 minutes.sherpa upreports a container as ready when it has started, so SSH answers a few minutes aftersherpa upfinishes. - With
ztp.method = "disabled"the node is the image as shipped: Sherpa mounts no config directory, so Junos boots from the image's own configuration. That configuration givesre0:mgmt-0the node's management IP address and turns on SSH, with only the image's ownadminuser: the node answers on its address, but it has nosherpalogin, sosherpa sshis refused. Log in asadminwith the password Juniper's documentation for the image gives. The node'svolumesfrom the manifest are mounted, so you can supply a configuration of your own.