ESI-LAG: Initial draft implementation

docs/plugins.md

@@ -21,6 +21,7 @@
    plugins/ospf.areas.md
    plugins/vrrp.version.md
    plugins/firewall.zonebased.md
+   plugins/evpn.multihoming.md
 ```
 
 Plugins needed by a topology file are listed in the **plugin** top-level element, for example:

docs/plugins/evpn.multihoming.md

@@ -0,0 +1,143 @@
+(plugin-evpn-multihoming)=
+# EVPN Multihoming (EVPN Ethernet Segment)
+
+This plugin allows for simple EVPN Ethernet Segment configuration, also know as EVPN Multihoming.
+
+For now, the plugin supports:
+* ES definition on LAG or "physical" interfaces (depending on platform support)
+* Only `all-active` multihoming
+* Manual or Auto-Generated (based on LACP) Ethernet Segment Identifier
+
+## Supported Platforms
+
+The plugin includes Jinja2 templates for the following platforms:
+
+| Operating system    | ES on LAG (ESI-LAG) | ES on other interfaces | Auto ESI |
+| ------------------- | :--: | :--: | :--: |
+| Arista EOS          |  ✅  |  ✅  |  ✅  |
+| Cumulus NVUE        |  ✅  |  ❌  |  ❌  |
+| vJunos Switch       |  ✅  |  ✅  |  ✅  |
+
+
+## Using the Plugin (auto mode)
+
+* Add `plugin: [ evpn.multihoming ]` to the lab topology.
+* Include the **evpn.es** attribute in the device interface
+
+netlab will generate, for each *Ethernet Segment*, ESI value (*Ethernet Segment ID*) and LACP System ID (for *ESI-LAG*).
+
+**NOTE**: Ethernet Segments ID will be generated starting from an integer value, which will be used as the first 5 most significant bytes (excluding the initial `0x00`). This is to:
+
+* be able to generate a 6-bytes LACP System ID starting with `0x02`.
+* be able to generate unique *ES-Import* target for each auto generated ESI value.
+
+## Using the Plugin (manual mode)
+
+It is also possible to manually define ESI values for your *Ethernet Segments*. In that case:
+
+* Add `plugin: [ evpn.multihoming ]` to the lab topology.
+* Define a set of *ethernet segments* on the topology top-level
+* Include the **evpn.es** attribute in the device interface
+
+**NOTE**: EVPN Multihoming, for ESI-LAG, requires that all LAG interfaces belonging to the same *ethernet segment* share the same (unique) LACP System ID. This can be achieved using the `lag.lacp_system_id` attribute - which can accept a "real" mac address value or an integer (*1-65535*) value: in that case it will generate a mac value in the format `02:xx:yy:xx:yy:00` (i.e., `1` will become `02:00:01:00:01:00`).
+
+### Supported attributes
+
+The plugin adds the following attributes defined at topology level:
+* **evpn.ethernet_segments** (dict) -- Key is the **ethernet segment** name. Each item is a *dict* with the following attributes:
+    * **id** (esi_id, mandatory) -- ESI in format `00:XX:XX:XX:XX:XX:XX:XX:XX:XX` (only Type-0 ESI is supported for now)
+    * **auto** (bool) -- Use ESI auto generation based on LACP System ID. If both `id` and `auto` are specified, explicit `id` takes over.
+
+Interface level attributes:
+* **evpn.es** (str) -- ethernet segment name (can be defined on `evpn.ethernet_segments`).
+
+## Example (auto mode)
+
+```
+plugin: [ 'evpn.multihoming' ]
+
+bgp.as: 65000
+
+groups:
+  _auto_create: true
+  switches:
+    members: [ s1, s2 ]
+    module: [ vlan, vxlan, ospf, bgp, evpn, lag ]
+  probes:
+    members: [ x1 ]
+    module: [ lag, vlan ]
+    device: eos
+  hosts:
+    members: [ h1, h2, h3 ]
+    device: linux
+    provider: clab
+
+vlans:
+  red:
+    mode: bridge
+    links: [ h1-x1, h2-s1, h3-s2 ]
+
+links:
+# EVPN/VXLAN Switch to Switch Link
+- s1:
+  s2:
+  mtu: 1600
+# ESI-LAG
+- lag:
+    members:
+    - s1:
+        evpn.es: seg_1
+      x1:
+    - s2:
+        evpn.es: seg_1
+      x1:
+  vlan.access: red
+```
+
+## Example (manual mode)
+
+```
+plugin: [ 'evpn.multihoming' ]
+
+bgp.as: 65000
+
+evpn.ethernet_segments:
+  seg_1.id: 00:11:22:33:44:55:66:77:88:99
+
+groups:
+  _auto_create: true
+  switches:
+    members: [ s1, s2 ]
+    module: [ vlan, vxlan, ospf, bgp, evpn, lag ]
+  probes:
+    members: [ x1 ]
+    module: [ lag, vlan ]
+    device: eos
+  hosts:
+    members: [ h1, h2, h3 ]
+    device: linux
+    provider: clab
+
+vlans:
+  red:
+    mode: bridge
+    links: [ h1-x1, h2-s1, h3-s2 ]
+
+links:
+# EVPN/VXLAN Switch to Switch Link
+- s1:
+  s2:
+  mtu: 1600
+# ESI-LAG
+- lag:
+    members:
+    - s1:
+        lag.lacp_system_id: 1
+        evpn.es: seg_1
+      x1:
+    - s2:
+        lag.lacp_system_id: 1
+        evpn.es: seg_1
+      x1:
+  vlan.access: red
+```

netsim/ansible/templates/lag/eos.j2

@@ -48,6 +48,10 @@ interface {{ intf.ifname }}
  description {{ intf.name }}
 {%  endif %}
 !
+{%  if intf.lag.lacp_system_id is defined %}
+ lacp system-id {{intf.lag.lacp_system_id}}
+{%  endif %}
+!
 {%   for ch in interfaces if ch.lag._parentindex|default(None) == intf.lag.ifindex %}
 !
 {%     set _lag_mode = 

netsim/ansible/templates/lag/junos.j2

@@ -25,9 +25,11 @@ interfaces {
         passive;
 {%     endif %}
         periodic {{intf.lag.lacp}};
+{%     if intf.lag.lacp_system_id is defined %}
+        system-id {{intf.lag.lacp_system_id}};
+{%     endif %}
       }
 {%   endif %}
-
     }
   }
 

netsim/ansible/templates/lag/linux.j2

@@ -19,6 +19,8 @@ ip link set dev {{ l.ifname }} up
 {%   endif %}
 {% endfor %}
 {% for l in interfaces if 'lag' in l and l.type in ['lag','bond'] %}
+{%   if l.lag.lacp_system_id is defined %}
+echo "{{ l.lag.lacp_system_id }}" > /sys/class/net/{{ l.ifname }}/bonding/ad_actor_system
+{%   endif %}
 ip link set dev {{ l.ifname }} up
 {% endfor %}
-exit 0

netsim/data/types.py

@@ -831,6 +831,22 @@ def must_be_mac(value: typing.Any) -> dict:
 
   return { '_valid': True }
 
+@type_test()
+def must_be_esi_value(value: typing.Any) -> dict:
+  # Allow only Type 0 ESI value to be supplied
+  _value_msg = '10-byte ESI Value in format 00:XX:XX:XX:XX:XX:XX:XX:XX:XX'
+
+  if not isinstance(value,str):
+    return { '_type': '10-byte ESI Value' }
+
+  if not value.startswith("00:"):
+    return { '_value': _value_msg + " - must start with '00:'" }
+
+  if not re.match(r"^00(:[0-9a-fA-F]{2}){9}$", value):
+    return { '_value': _value_msg }
+
+  return { '_valid': True }
+
 @type_test()
 def must_be_net(value: typing.Any) -> dict:
   if not isinstance(value,str):

netsim/devices/eos.py

@@ -101,6 +101,13 @@ def passive_stub_interfaces(node: Box, topology: Box) -> None:
 
     intf.ospf.network_type = 'point-to-point'
 
+def esi_identifier_format(node: Box, topology: Box) -> None:
+  # Arista EOS, in EVPN ESI, wants ESI in format: 0000:0000:0000:0000:0000
+  for intf in node.interfaces:
+    if intf.get('evpn._esi.id'):
+      esi = intf.evpn._esi.id.replace(":", "")
+      intf.evpn._esi._eos_id = ':'.join(esi[i:i+4] for i in range(0, len(esi), 4))
+
 class EOS(_Quirks):
 
   @classmethod
@@ -118,3 +125,5 @@ def device_quirks(self, node: Box, topology: Box) -> None:
       passive_stub_interfaces(node,topology)
     if 'eos' in node:
       configure_ceos_attributes(node,topology)
+    if 'evpn.multihoming' in topology.get('plugin',[]):
+      esi_identifier_format(node,topology)

netsim/extra/evpn.multihoming/cumulus_nvue.j2

@@ -0,0 +1,20 @@
+{% for i in interfaces if i.type == 'lag' and i.evpn._esi is defined %}
+{%  if loop.first %}
+- set:
+    evpn:
+      enable: on
+      multihoming:
+        enable: on
+        startup-delay: 1
+        mac-holdtime: 30
+        neighbor-holdtime: 30
+    interface:
+{%  endif %}
+      {{ i.ifname }}:
+        evpn:
+          multihoming:
+            segment:
+              enable: on
+              mac-address: {{i.lag.lacp_system_id}}
+              identifier: {{i.evpn._esi.id}}
+{% endfor %}

netsim/extra/evpn.multihoming/defaults.yml

@@ -0,0 +1,55 @@
+# evpn.multihoming: Support EVPN Multihoming (Ethernet Segments)
+#
+---
+devices:
+  vjunos-switch:
+    features.evpn.multihoming:
+      # support ES definition on LAG interfaces (ESI-LAG)
+      lag: true
+      # support ES definition on "standard" ethernet interfaces
+      interface: true
+      # ESI Auto value based on LACP System ID
+      esi_auto: true
+      modes: [ 'all-active' ] # not used yet
+  eos:
+    features.evpn.multihoming:
+      lag: true
+      interface: true
+      esi_auto: true
+      modes: [ 'all-active' ] # not used yet
+  cumulus_nvue:
+    features.evpn.multihoming:
+      lag: true
+      interface: false
+      esi_auto: false
+      modes: [ 'all-active' ] # not used yet
+  # for validation
+  none:
+    features.evpn.multihoming:
+      # support ES definition on LAG interfaces (ESI-LAG)
+      lag: true
+      # support ES definition on "standard" ethernet interfaces
+      interface: true
+      # ESI Auto value based on LACP System ID
+      esi_auto: true
+      modes: [ 'all-active' ] # not used yet
+
+
+# data model:
+# -- global: evpn.ethernet_segments
+# -- on interface: es: xxx (must be a valid ethernet_segment reference)
+evpn:
+  attributes:
+    global:
+      ethernet_segments:
+        type: dict
+        _keytype: id
+        _subtype: _es_item
+    node:
+      ethernet_segments: { copy: global }
+    interface:
+      es: id
+    # specific attributes
+    _es_item:
+      auto: bool
+      id: esi_value

netsim/extra/evpn.multihoming/eos.j2

@@ -0,0 +1,15 @@
+{% for intf in interfaces if intf.evpn._esi is defined %}
+interface {{ intf.ifname }}
+ evpn ethernet-segment
+{%   if intf.evpn._esi.auto|default(false) %}
+  identifier auto lacp
+{%   else %}
+{# Arista wants it in format: 0000:0000:0000:0000:0000 (generated by device quirk) #}
+  identifier {{intf.evpn._esi._eos_id}}
+{%   endif %}
+!
+{% endfor %}
+!
+router bgp {{ bgp.as }}
+ address-family evpn
+  route type ethernet-segment route-target auto

netsim/extra/evpn.multihoming/frr.j2

@@ -0,0 +1,18 @@
+#!/bin/bash
+
+{### STILL NOT WORKING - to be better investigated ###}
+
+# ESI-LAG config in FRR
+cat >/tmp/esi_lag_config <<CONFIG
+evpn mh startup-delay 1
+!
+{% for intf in interfaces if intf.type == 'lag' and intf.evpn._esi is defined %}
+interface {{ intf.ifname }}
+  evpn mh es-id {{intf.evpn._esi.id}}
+!
+{% endfor %}
+!
+do write
+CONFIG
+vtysh -f /tmp/esi_lag_config
+vtysh -c 'clear bgp *'