Cumulus_NVUE: Add mlag support

docs/module/lag.md

@@ -83,7 +83,8 @@ The above diagram illustrates the 3 supported topologies:
 * 2:2 dual mlag between 2 pairs of nodes (4 nodes in total)
 
 MLAG related parameters:
-* **lag.mlag.peerlink**: Used on peerlink to configure a unique ID for the pair of switches providing the MLAG. Can be set to *True* for auto-id generation, or an integer (that must be globally unique)
+* **lag.mlag.mac**: Used at node or interface level to configure the MAC address for the peerlink. *Netlab* can auto-generate this, so it is normally not necessary to set this
+* **lag.mlag.peergroup**: Used on peerlink to configure a unique ID for the pair of switches providing the MLAG. Can be set to *True* for auto-id generation, or an integer (that must be globally unique)
 
 A simple example:
 ```
@@ -94,7 +95,7 @@ groups:
   device: dellos10
  hosts:
   members: [h1,h2]
-  device: frr          # 'linux' does not support the lag module yet
+  device: frr           # 'linux' does not support the lag module yet
 
 links:
 - lag:
@@ -104,5 +105,5 @@ links:
 
 # Inter-switch peer link(s) for MLAG sync
 - lag:
-   members: [s1-s2]    # Note that multiple physical links are allowed here
-   mlag.peerlink: True # (also) used to derive a unique MAC address for this group of MLAG peers
+   members: [s1-s2]     # Note that multiple physical links are allowed here
+   mlag.peergroup: True # (also) used to derive a unique MAC address for this group of MLAG peers

netsim/ansible/templates/lag/cumulus_nvue.j2

@@ -1,10 +1,40 @@
+{% if lag.mlag is defined %}
+- set:
+    interface:
+      peerlink:
+        bond:
+          member:
+{% for intf in interfaces if intf.lag.mlag.peergroup is defined %}
+            {{ intf.ifname }}: {}
+{%   for ch in interfaces if ch.lag._parentindex|default(False) == intf.linkindex %}
+            {{ ch.ifname }}: {}
+{%   endfor %}
+{% endfor %}
+        type: peerlink
+      peerlink.{{ lag.mlag.vlan }}:
+        base-interface: peerlink
+        type: sub
+        vlan: {{ lag.mlag.vlan }}
+    mlag:
+      backup:
+        {{ lag.mlag.peer_backup_ip }}: {}
+      enable: on
+      mac-address: {{ lag.mlag.mac | hwaddr('linux') }}
+      peer-ip: {{ lag.mlag.peer }}
+{% endif %}
+
 {% for i in interfaces if i.type == 'lag' %}
 {%  if loop.first %}
 - set:
     interface:
 {%  endif %}
       {{ i.ifname }}:
         bond:
+{%  if '_mlag' in i.lag %}
+          mlag:
+            enable: on
+            id: {{ i.lag.ifindex }}
+{%  endif %}
 {%  set _lacp = i.lag.lacp|default(lag.lacp) %}
 {%  if _lacp=='slow' %}
           lacp-rate: slow

netsim/devices/cumulus_nvue.yml

@@ -31,6 +31,13 @@ features:
     activate_af: True
   lag:
     passive: False
+    mlag:
+      peer:
+        global: True              # Use global node level settings for peerlink
+        mac: 44:38:39:ff:00:00    # Base to generate a virtual MAC. NVidia reserved range
+        vlan: 4094                # Use this vlan
+        ip: linklocal             # Use this IP subnet
+        backup_ip: loopback.ipv4  # Use loopback.ipv4 as a backup IP address for peerlink
   ospf:
     unnumbered: True
   stp:

netsim/modules/lag.py

@@ -325,33 +325,44 @@ def process_lag_links(topology: Box) -> None:
 # populate_mlag_peer - Lookup the IPv4 loopback address for the mlag peer, and derive a virtual MAC to use
 #
 def populate_mlag_peer(node: Box, intf: Box, topology: Box) -> None:
-  _n = intf.neighbors[0].node
-  peer = topology.nodes[_n]
+  peer = topology.nodes[intf.neighbors[0].node]
   features = devices.get_device_features(node,topology.defaults)
-  _mlag_peer = features.get('lag.mlag.peer',{})
-  if 'ip' in _mlag_peer:
-    if 'loopback' in _mlag_peer.ip:
-      _ip = peer.get(_mlag_peer.ip,None)
-      if _ip:
-        intf.lag.mlag.peer = str(netaddr.IPNetwork(_ip).ip)
+  mlag_peer = features.get('lag.mlag.peer',{})
+  _target = node.lag if mlag_peer.get('global',False) else intf.lag         # Set at node or intf level?
+  if 'ip' in mlag_peer:
+    if mlag_peer.ip == 'linklocal':
+      _target.mlag.peer = 'linklocal'
+    elif 'loopback' in mlag_peer.ip:                                        # Could check if an IGP is configured
+      ip = peer.get(mlag_peer.ip,None)
+      if ip:
+        _target.mlag.peer = str(netaddr.IPNetwork(ip).ip)
       else:
-        log.error(f'Node {peer.name} must have {_mlag_peer.ip} defined to support MLAG',
+        log.error(f'Node {peer.name} must have {mlag_peer.ip} defined to support MLAG',
           category=log.IncorrectValue,
           module='lag')
     else:
-      _net = netaddr.IPNetwork(_mlag_peer.ip)
-      _id = 0 if node.id < peer.id else 1
-      intf.lag.mlag.peer = str(_net[_id])
-      intf.lag.mlag.self = f"{_net[1-_id]}/{_net.prefixlen}"                # including /prefix
+      net = netaddr.IPNetwork(mlag_peer.ip)
+      id = 0 if node.id < peer.id else 1
+      _target.mlag.peer = str(net[1-id])                                    # Higher node ID gets .1
+      _target.mlag.self = f"{net[id]}/{net.prefixlen}"                      # including /prefix
+
+  if 'backup_ip' in mlag_peer:
+    bk_ip = peer.get(mlag_peer.backup_ip,None)
+    if bk_ip:
+      _target.mlag.peer_backup_ip = str(netaddr.IPNetwork(bk_ip).ip)
+    else:
+      log.error(f'Node {peer.name} must have "{mlag_peer.backup_ip}" defined to support backup MLAG peerlink',
+                category=log.IncorrectValue,
+                module='lag')
 
-  if 'mac' in _mlag_peer:
-    _mac = netaddr.EUI(_mlag_peer.mac)                                      # Generate unique virtual MAC per MLAG group
-    _mac._set_value(_mac.value + intf.get(PEERLINK_ID_ATT,0) % 65536 )      # ...based on lag.mlag.peergroup
-    intf.lag.mlag.mac = str(_mac)
+  if 'mac' in mlag_peer and not isinstance(_target.get('mlag.mac',None),str):
+    mac = netaddr.EUI(mlag_peer.mac)                                        # Generate unique virtual MAC per MLAG group
+    mac._set_value(mac.value + intf.get(PEERLINK_ID_ATT,0) % 65536 )        # ...based on lag.mlag.peergroup
+    _target.mlag.mac = str(mac)
 
   for v in ['vlan','ifindex']:
-    if v in _mlag_peer:
-      intf.lag.mlag[v] = _mlag_peer[v]
+    if v in mlag_peer:
+      _target.mlag[v] = mlag_peer[v]
 
   intf.pop('vlan',None)                                                     # Remove any VLANs provisioned on peerlinks
 
@@ -383,11 +394,11 @@ def node_post_transform(self, node: Box, topology: Box) -> None:
         populate_mlag_peer(node,i,topology)
         has_peerlink = True
       elif i.type=='lag':
-        i.lag = node.get('lag',{}) + i.lag  # Merge node level settings with interface overrides
-        # i.pop('mtu',None)                 # Next PR: Remove any MTU settings - inherited from members
-        lacp_mode = i.get('lag.lacp_mode')  # Inheritance copying is done elsewhere
+        node_atts = { k:v for k,v in node.get('lag',{}).items() if k!='mlag'}
+        i.lag = node_atts + i.lag           # Merge node level settings with interface overrides
+        lacp_mode = i.get('lag.lacp_mode')
         if lacp_mode=='passive' and not features.lag.get('passive',False):
-          log.error(f'Node {node.name} does not support passive LACP configured on interface {i.ifname}',
+          log.error(f'Node {node.name}({node.device}) does not support passive LACP configured on interface {i.ifname}',
             category=log.IncorrectAttr,
             module='lag')
         if i.lag.get('_mlag',False) is True:

netsim/modules/lag.yml

@@ -19,6 +19,8 @@ attributes:
     lacp:
     lacp_mode:
     mode:
+    mlag:
+      mac: mac                  # MAC to use for MLAG peering, auto-derived if not set
   link:                         # Most should be consistent across both interfaces on the link
     lacp: { copy: global }
     lacp_mode: { copy: global }

tests/integration/lag/09-mlag-m-to-m.yml

@@ -8,7 +8,7 @@ groups:
   _auto_create: true
   switches:
     members: [a1, a2, b1, b2]
-    module: [lag, vlan]
+    module: [ lag, vlan ]     # May need to add OSPF for backup peerlink reachability
   hosts:
     members: [h1, h2]
     device: linux
@@ -36,8 +36,8 @@ validate:
   ping:
     description: Pinging H2 from H1
     nodes: [h1]
-    wait_msg: Waiting for STP to enable the ports
-    wait: 45
+    wait_msg: Waiting for STP and MLAG init-delay timer to enable the ports
+    wait: 180
     plugin: ping('h2')
   ping_gw:
     description: Pinging gateway from H1

tests/topology/expected/lag-mlag-m_to_m.yml

@@ -209,9 +209,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.0
+          peer: 169.254.127.1
           peergroup: 1
-          self: 169.254.127.1/31
+          self: 169.254.127.0/31
           vlan: 4094
       linkindex: 1
       name: a1 -> a2
@@ -330,9 +330,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.1
+          peer: 169.254.127.0
           peergroup: 1
-          self: 169.254.127.0/31
+          self: 169.254.127.1/31
           vlan: 4094
       linkindex: 1
       name: a2 -> a1
@@ -438,9 +438,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.0
+          peer: 169.254.127.1
           peergroup: 2
-          self: 169.254.127.1/31
+          self: 169.254.127.0/31
           vlan: 4094
       linkindex: 2
       name: b1 -> b2
@@ -546,9 +546,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.1
+          peer: 169.254.127.0
           peergroup: 2
-          self: 169.254.127.0/31
+          self: 169.254.127.1/31
           vlan: 4094
       linkindex: 2
       name: b2 -> b1

tests/topology/expected/node.clone-plugin-lag.yml

@@ -747,9 +747,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.0
+          peer: 169.254.127.1
           peergroup: 1
-          self: 169.254.127.1/31
+          self: 169.254.127.0/31
           vlan: 4094
       linkindex: 1
       name: r1 -> r2
@@ -953,9 +953,9 @@ nodes:
       lag:
         mlag:
           ifindex: 4094
-          peer: 169.254.127.1
+          peer: 169.254.127.0
           peergroup: 1
-          self: 169.254.127.0/31
+          self: 169.254.127.1/31
           vlan: 4094
       linkindex: 1
       name: r2 -> r1