Cumulus Linux has a cool ‘feature’ it inherited from Linux. On most network switches a port is either Layer2 or Layer3. When a port is layer2 it has to be part of a VLAN. Linux does understand and work with VLANs but it can have a port that is running at layer2 but not part of a VLAN. We can configure a port under \/etc\/network\/interfaces like this:<\/p>\n
auto swp1\r\niface swp1\r\n<\/pre>\nIf we ifup this port its not part of a VLAN, and its not Layer3. <\/p>\n
Why would we want to do this? Honestly the most common reason is that we can check physical connections by using lldp before this port is part of a broadcast domain (VLAN) that could cause loops or unexpected behavior.<\/p>\n
What if I cabled a switch and don’t know what ports are connected? You could create a stanza for each swp like above, use something like mako<\/a> or do a simple bash loop on the command line like this:<\/p>\n
cumulus@leaf01:~$ for swp in {1..54}; do sudo ip link set swp$swp; done\r\n<\/pre>\nNow all 54 ports are admin up at layer2 so we can check connections, but its not routing or switching. Now you can use a “net show int” or use the linux command “ip link show”<\/p>\n
cumulus@leaf01:~$ ip link show | grep LOWER_UP\r\n1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default\r\n2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master mgmt state UP mode DEFAULT group default qlen 1000\r\n11: swp9: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9216 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000\r\n51: swp49: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 9216 qdisc pfifo_fast master peerlink state UP mode DEFAULT group default qlen 1000\r\n52: swp50: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 9216 qdisc pfifo_fast master peerlink state UP mode DEFAULT group default qlen 1000\r\n53: swp51: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9216 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000\r\n54: swp52: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9216 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000<\/pre>\nYou can even just use lldp here<\/p>\n
cumulus@leaf01:~$ net show lldp\r\n\r\nLocal Port Speed Mode Remote Port Remote Host Summary\r\n------------ ------- ------------- ---- ------------- --------------- --------------------------\r\neth0 1G Mgmt ==== swp23 oob-mgmt-switch IP: 10.50.100.100\/24(DHCP)\r\nswp9 10G NotConfigured ==== swp51 leaf07\r\nswp49 40G NotConfigured ==== swp49 exit02 \r\nswp50 40G NotConfigured ==== swp50 exit02 \r\nswp51 40G NotConfigured ==== swp30 spine01\r\nswp52 40G NotConfigured ==== swp30 spine02\r\n<\/pre>\nYay now we can go configure them now that we know how they are cabled up. Its great to be lazy \ud83d\ude42<\/p>\n
Update 3\/30\/2017<\/p>\n
The famous Daniel Walton (yeah the Daniel that has his name on this<\/a>, this<\/a> and my personal favorite this<\/a>.) let me know that Cumulus Linux’s new NCLU (Network Command Line Utility) also has a method of doing this quickly. You can check out his tweet here.<\/a> NCLU uses the net command and has the ability to do a range of ports really quickly. One caveat vs using ip link set… this will make persistent config (meaning the ports are actually configured under \/etc\/network\/interfaces). If you don’t want the config to remain you can do a “net del” then a “net commit”<\/p>\n
\r\nnet add int swp1-54\r\n<\/pre>\n","protected":false},"excerpt":{"rendered":"Cumulus Linux has a cool ‘feature’ it inherited from Linux. On most network switches a port is either Layer2 or Layer3. When a port is layer2 it has to be part of a VLAN. Linux does understand and work with VLANs but it can have a port that is running at layer2 but not part of a … <\/p>\n