IP fabric over unnumbered interfaces

  So, you read the industry websites, know that IP fabric is the next best thing in data center networking and decided to take a plunge and build your own. Nothing big to start with: 2 spine and 8 leaf switches. Now you realize that your IPAM system does not have API and you have to assign 16 IP addresses for transit links and 10 IPs for loopback interfaces manually. While not insurmountable task, it's tedious. Fortunately, Cisco's NXOS and Juniper's JUNOS let you configure ip unnumbered ethernet interface and now you need only 10 IPs for loopbacks.
  I created a small virtual lab of 3 NXOSv switches - 2 spines and 2 leafs - to test the concept. I could not make BGP work directly over unnumbered interfaces, so I configured OSPF to advertise loopbacks and BGP peering between loopbacks. To simplify configuration even more, I configured dynamic BGP peering on the spines.
   Why would you need BGP if you already have OSPF? You might want to run another next best thing - VXLAN with EVPN control plane.
  To run the lab you'll need Vagrant, VirtualBox, Linux machine with 32GB of memory and Vagrant package of NXOS from Cisco. You need to have CCO account and may be maintenance contract to download the image. Do not ask me to provide the image. I used NXOSv image nxosv-final.7.0.3.I7.1.box, Vagrant 2.0.1, VirtualBox 5.2.6 and Ubuntu Linux 16.04LTS. Although the lab worked with earlier versions of Vagrant and VirtualBox and should run on any Linux distro.

• Clone or download from git configuration files
• Run "create_iso.sh" script to build ISO files with configuration for each NXOSv switch
• run "vagrant up". Depending on the resources, it might take up 10 minutes for all 4 switches to come up fully.

You can see boot progress by connecting to consoles: "ncat -U /tmp/", where are leaf1, leaf2, spine1 or spine2. After switches are up, you can log in by running "vagrant ssh ". You'll be presented with bash shell, to get to NXOS prompt, type "su - admin", password is admin.

Disclaimer: this is in no way shape or form production-ready configuration and was not tested for any side effects. Use it at your own risk.

Happy Labbing!

Cattle vs Pets rant

Ok, by now pretty much everybody in IT knows about cattle and pets meme. Your IT infrastructure should be disposable cattle and not precious pets. While I agree that IT infrastructure should be disposable, I have a problem with this specific choice of words. Only people who never set foot in a cattle ranch can say that. Any farm roaming bovine actually brings money, so treating one dead cow like it's nothing to worry about is going to cost you. One sick cow must get as much, if not more, attention than pet cat or your entire operations can be in jeopardy.
Let's come up with different analogy.

It does not take that long

When some overly enthusiastic SDN neophyte or sysadmin-turned-devops-engineer tells me about how long it used to take to create a VLAN and how quick it can be done with new magic SDN controller or automation tool of the day, I have to assume that this person does not really know what she or he is talking about.
What takes long time is change control procedure and no amount of automation or SDN is going to change it.
Since we are talking about VLANs here, in modern data center network you do not need to create a VLAN on more than one switch.

Manage network devices with Ansible

Ansible is one of the best technologies we took from buggers after the war.  I loved "Ender's game" book, not the movie.

Inspired by excellent posts  by Kirk Byers I decided to try Ansible not only to generate configuration for network switches, but to make configuration changes. I have virtual Arista switch running in VirtualBox, so this is where I ran my tests, but it's easy to replicate with Juniper or Cisco Nexus switches. I used user "root", although any user with priviledge level 15 will do.

First, enable root user on Arista switch:

Arista-5#(conf) aaa root secret SecretPassword

Next step is to go to managemnent server and generate ssh key without password. Resulted public key should be added to /root/authorized_keys file on Arista switch.

Now, to Ansible.

My ansible.cfg:

[defaults]

host_key_checking=False

hostfile=/home/user1/ansible/hosts

log_path=~/ansible.log

Let's do very simple task: copy new OS image file and update boot variable. Here is my very simple playbook upgrade.yml:

---
- hosts: arista
  remote_user: root
  tasks:
  - name: Push image
     copy: src=/home/user1/Documents/ansible/vEOS-1.swi dest=/mnt/flash/vEOS-1.swi
  - name: Change boot variable
    command: FastCli -p15 -c "install source vEOS-1.swi now"

Really simple inventory file:

[arista]

arista-5

Let's run it:

It worked, boot variable now points to vEOS-1.swi file.

What happens if you you use RADIUS for authentication and have to enter 

password to log into your switch? In this case Ansible uses sshpass which stores your password in the memory. From sshpass man page:

It is close to impossible to securely store the password, and users of sshpass should consider whether ssh's public key authentication provides the same end-user experience, while involving less hassle and being more secure.

If you are willing to take this risk, insert "ask_pass=True" line into your ansible.cfg to be prompted for password or run ansible_playbook command with -k option.

Before you finish your next automation script

An excellent article from Wall Street Journal about pitfalls of automation:

This philosophy traps people in a vicious cycle of de-skilling. By isolating them from hard work, it dulls their skills and increases the odds that they will make mistakes. When those mistakes happen, designers respond by seeking to further restrict people’s responsibilities—spurring a new round of de-skilling.

Something to consider when you hand over that new automation script to your network operations center.

Al-too-na! Al-too-na!

There is a lot of press these days about new Facebook datacenter and how it (almost) means death of Cisco-Juniper-Arista. Here are the questions:
Did somebody try to run multicast on white-box switch?
Where do you place Rendezvous Point in this network?
There is only handful of companies with Facebook- or Google-size network, but hundreds of companies requiring multicast support in their data center. Until white box switches learn to run IGMP/PIM/MSDP Cisco and Juniper will have plenty of customers with deep pockets.

Auto Provisioning

Network automation is hot topic lately and has been my favorite past time for quite a while. While vendors provided tools to make changes on the switches and routers - Cisco Works come to mind - getting initial configuration on the switch was still manual process. Network operators would come up with some kind of template in Excel or text file and then replace host name and management IP address. To get initial configuration on the switch, one would use copy/paste, a process prone to errors, or upload text file to the switch and copy it to startup-config or candidate configuration. Latter process required configured management IP address and default gateway.  In addition to that operators had to upgrade/downgrade operating system.
Fast forward to today's era of Chef/Puppet/Ansible/Other appropriated household word. Cisco came up with (and Arista duly replicated) Power-on Auto Provisioning for its Nexus switches, POAP for short. In DHCP option 67 you need to provide Python script name that switch will download and execute. Arista also supports Bash scripts. Cisco even provides a script that can upgrade NX-OS and download corresponding configuration file based on one of the following parameters: switch name, management interface MAC address, serial number or CDP neighbors. You can write your own script to generate configuration on the fly.
Juniper came up with somewhat catchier name - Zero Touch Provisioning. It does not allow you run scripts. Juniper touted Junos as first network operating system that allowed scripts long before Cisco came up with EEM. Also, you have to use DHCP vendor options to encode configuration file name and Junos image file name.  So, in your DHCP server you have to put something like this: 0x00306a696e7374616c6c2d7166782d332d31332e325835302d4431352e332d646f6d65737469
632d7369676e65642e74677a0111646e6a722d6c61622d716678312e636667. Not very informative. Juniper promised to implement SLAX script support in near future and abandon DHCP vendor option in favor of option 67. Python support might come later.

Python on Nexus

If you are relying on handy "cisco" python module on your Nexus 5500 or 6K switches - stop. Cisco made considerable changes in version 7.0(1)N1(1).
Here what it looked like in 6.0(2)N2(3):
switch# python
Python 2.7.2 (default, Nov 27 2012, 17:50:33)
[GCC 4.3.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
Loaded cisco NxOS lib!
>>> import cisco
>>> dir(cisco)
['BGPSession', 'BufferDepthMonitor', 'CLI', 'CheckPortDiscards', 'CiscoSecret', 'CiscoSocket', 'Feature', 'History', 'IPv4ACL', 'IPv6ACL', 'Interface', 'Key', 'LineParser', 'MacAddressTable', 'OSPFSession', 'Routes', 'SectionParser', 'VRF', 'Vlan', '__all__', '__builtins__', '__doc__', '__file__', '__name__', '__package__', '__path__', 'acl', 'bfd', 'bgp', 'buffer_depth_monitor', 'check_port_discards', 'cisco_secret', 'cisco_socket', 'cli', 'dhcp', 'eigrp', 'feature', 'get_global_vrf', 'get_valid_port', 'history', 'hsrp', 'interface', 'interface-vlan', 'key', 'lacp', 'line_parser', 'mac_address_table', 'md5sum', 'msdp', 'ospf', 'ospfv3', 'pim', 'private-vlan', 'ptp', 'rip', 'routes', 'scheduler', 'section_parser', 'set_global_vrf', 'show_queues', 'show_run', 'ssh', 'tacacs', 'telnet', 'transfer', 'udld', 'vlan', 'vpc', 'vrf', 'vrrp', 'vtp']
>>>

In 7.0(1)N1(1)
>>> import cisco
>>> dir(cisco)
['__doc__', '__name__', '__package__', 'cli', 'cli_execution_error', 'cli_syntax_error', 'clid', 'clip']

According to Cisco representative "Starting 7.0(1)N1(1), python interpreter on N6K has been modified to look like python interpreter on N7K."
WHY DID NOT THEY DO OTHER WAY AROUND: MODIFY PYTHON ON N7K TO MATCH N6K?

Stop using "ships in the night"

It seems that every month brings news of yet another overlay network. STT, OTV, VXLAN, NVGRE and may be many more I've not heard of. People often use "ships in the night" words in the same sentence referring to the fact that these overlay network know nothing about underlying physical network infrastructure. However, nowadays even the smallest sea or ocean going vessel not only knows about any ship nearby, but acutely aware of a few satellites thanks to radar and GPS systems.
So, to keep up with pace of time and advances in technologies (after all, we are in tech business) I propose  new analogy for overlay and physical networks: polar bear and penguin. Those 2 definitely do not meet in real life. I am not considering corner case like zoo.

Rant

Any vendor requiring registration in order to access online documentation should be condemned to untangle network cables for the rest of the their product life.

Very interesting and popular explanation of Shennon limit: Part 1 and Part 2.

LACP timer and what it means

LACP (IEEE 802.3ad)is protocol used to bundle several physical interfaces to form single logical channel. It has a timer which defines how often devices inter-connected via this bundle exchange LACP PDUs or control messages. Currently, this timer can be set to either "rate fast" - 1 second, or "rate normal" - 30 seconds. What is not always clear is that when you configure "lacp rate " on Cisco or "set interfaces ae1 aggregated-ether-options lacp periodic fast" on Juniper, you do not configure how often this switch will send LACP PDUs. This command means that switch where this command is applied will expect to receive LACP PDUs with this frequency from the partner on the other side of logical channel.
Here is quick test. I have Nexus5500 connected to Cat6500. Let's configure port-channel between them with one physical member interface.

Cat6500#show run interface TenGigabitEthernet 1/5
!
interface TenGigabitEthernet1/5
 switchport
 switchport trunk encapsulation dot1q
 switchport mode trunk
 lacp rate fast
 channel-group 5 mode active

Nexus5500# show running-config interface Ethernet 1/1
!
interface Ethernet1/1
  switchport mode trunk
  channel-group 1 mode active

"lacp rate normal" is default setting on Nexus, so this command does not show up in the output, but we can confirm:

Nexus5500# show running-config interface Ethernet 1/1 all | include lacp
  lacp port-priority 32768
  lacp rate normal


Cat6500 is configured with rate fast and Nexus5500 - with rate normal. Let's see what's going on behind the scene.
On Catalyst:
Cat6500#show lacp internal
Flags:  S - Device is requesting Slow LACPDUs
           F - Device is requesting Fast LACPDUs
           A - Device is in Active mode       P - Device is in Passive mode    

Channel group 3
                            LACP port     Admin     Oper    Port        Port
Port      Flags   State     Priority      Key       Key     Number      State
Te1/5     FA      bndl      32768         0x3       0x3     0x106       0x3F

F flags says that Cat6500 requesting fast LACP PDUs from its partner.

On Nexus it's a little bit backwards, the "show" command tells you partner status, not its own.

Nexus5500# show lacp neighbor interface port-channel 1
Flags:  S - Device is sending Slow LACPDUs F - Device is sending Fast LACPDUs
          A - Device is in Active mode       P - Device is in Passive mode
port-channel1 neighbors
Partner's information
               Partner                           Partner                            Partner
Port        System ID                      Port Number     Age         Flags
Eth1/1     20,0-13-5f-20-63-80    0x106               910         SA

            LACP Partner           Partner                     Partner
            Port Priority              Oper Key                 Port State
            32768                      0x3                         0x3f

Nexus5500 says, that its partner - Cat6500 - is sending LACP PDUs every 30 seconds.

Find MAC address for IPv4 Multicast group

When troubleshooting multicast problem I find myself checking if IGMP snooping works as intended. "show mac address-table multicast" on Cisco switches shows MAC addresses of multicast groups. Tired of converting Multicast IPs into MACs with pencil and paper, I wrote my first ever script in Python which does just that. It takes IP address of multicast group as a parameter. Although I did some testing there might be bugs, so beware.
See Cisco's white paper for explanation how the conversion is done.
 

The wait is over

Finally, in NX-OS 6.0(2) for Nexus 5000 platform Cisco implemented "default interface" command which lets you return interface to its factory default configuration. It is very-very-very useful feature in the lab environment when one has to do a lot of re-configuration and something does not work as expected simply because of left-over configuration from the previous test. 

This command has been available in IOS since 11.1 and in NX-OS for Nexus 7K since 5.1(1)

Juniper haiku

No wonder Juniper software is so bloated :).
admin@switch> show version and haiku
Hostname: switch
Model: ex4200-48t
JUNOS Base OS boot [10.4R12.1]
JUNOS Base OS Software Suite [10.4R12.1]
JUNOS Kernel Software Suite [10.4R12.1]
JUNOS Crypto Software Suite [10.4R12.1]
JUNOS Online Documentation [10.4R12.1]
JUNOS Enterprise Software Suite [10.4R12.1]
JUNOS Packet Forwarding Engine Enterprise Software Suite [10.4R12.1]
JUNOS Routing Software Suite [10.4R12.1]
JUNOS Web Management [10.4R12.1]


        Now that Zion's safe,
        can they find a better place
        to get some sweaters?

admin@switch> show version and haiku

        Juniper babies
        The next generation starts
        Gotta get more sleep

admin@switch> show version and haiku

        3am; darkness;
        Maintenance window closing.
        Safety net: rollback.

admin@switch> show version and haiku

        Shiny leather pants
        Why don't they squeak when they kick?
        Is _that_ the secret?

PACL and MAC address learning

I have unenviable task to drag legacy application to 21st century. I am talking about 80-the legacy and some of its functions do not even use IP protocols. One of the proposed solution included 2 servers with the same

IP and MAC addresses (I know, but splitting network in 2 separate VLANs was not an option) connected to different switches, but in the same VLAN. ClientA and ClientB should be able to talk to each other and server connected to the same switch as client. ServerA and ServerB should not even know about each other's existence, so they won't complain about duplicate IP address. The switches are Cisco 6500s. One of the obvious  solutions is to put Port Access Control List on either side of the inter-switch link. PACL successfully blocked the traffic between servers, but switches still learned MAC address of the blocked traffic source and placed it MAC address table. This would cause MAC address flapping on the switch every time clients send ARP query for server's MAC or when both servers need to send traffic to their clients. Why would switch need to keep MAC address of the discarded traffic? Oh, well. Another network mystery.

"private-vlan syncronize" pitfall

Let's say you need to configure Private VLAN on Cisco Nexus switch with MSTP. You create new VLAN, make it isolated and map it to primary VLAN. Your newly created VLAN is automatically mapped to MST0 and if your primary happened to be in any other MST instance there is a possibility that your primary and secondary VLANs ended up with different L2 paths. As a bonus, you get
annoying message "These secondary vlans are not mapped to the same instance as their primary" every time you run "show spanning-tree configuration". Not a big deal, but things like these irritate me. Not to worry, "private-vlan syncronize" under "spanning-tree mst configuration" will automatically map all secondary VLANs to the same MST instance as primary VLAN. The moment it's done, you MST digest changes and boom, you have brand new MST region and STP convergence on top of it. So, either pick exiting VLAN mapped to the same instance and convert to secondary community or isolated VLAN, or have all your VLANs mapped to MST0 and use other methods to load-share traffic between transit links.
I know what you are thinking. No, I did it in the lab, so you won't have to.

How complex systems fail

A must read paper for every system architect. Also recording of the talk given at Velocity conference by Richard Cook. The main take-away for me is: don't build reliable system, build resilient system.

Port-security side effect

I discovered interesting side effect of configuring port-security which can prevent its deployment in certain circumstances. Let's have a look. Here is part of pertaining interface configuration:

 switchport port-security maximum 20
 switchport port-security
 switchport port-security aging time 1440
 switchport port-security violation restrict
 switchport port-security aging type inactivity

You can find explanations of what each command does here.

switch# show port-security interface gi 0/45
Port Security                      : Enabled
Port Status                          : Secure-up
Violation Mode                  : Restrict
Aging Time                        : 1440 mins
Aging Type                        : Inactivity
SecureStatic Address Aging : Disabled
Maximum MAC Addresses      : 20
Total MAC Addresses              : 1
Configured MAC Addresses    : 0
Sticky MAC Addresses            : 0
Last Source Address:Vlan        : abcd.ef12.3456:1234
Security Violation Count          : 0

In case you were wondering, the MAC address above is completely made up, but it associated with "floating" IP assigned to active device in cluster. Let's see what happens when active IP has to be moved to other device in cluster due to fail-over:

port_security-2-psecure_violation: security violation occurred, caused by mac address abcd.ef12.3456 on port gigabitethernet0/46.
Oops. Even though port-security configuration allows to learn up to 20 MAC addresses and there are no MAC addresses on Gi0/46, we got port-security violation. Why? Let's see debug port-security:


 PSECURE: psecure_add_addr_check: Found duplicate mac-address abcd.ef12.3456, It is already secured on Gi0/45
%PORT_SECURITY-2-PSECURE_VIOLATION: Security violation occurred, caused by MAC address abcd.ef12.3456 on port GigabitEthernet0/46.
PSECURE: Security violation, TrapCount:346
PSECURE: Read:2830, Write:2831
PSECURE: swidb = GigabitEthernet0/46 mac_addr = abcd.ef12.3456 vlanid = 1234
PSECURE: Adding abcd.ef12.3456 as dynamic on port Gi0/46 for vlan 1234
PSECURE: Violation/duplicate detected upon receiving abcd.ef12.3456 on vlan 1234: port_num_addrs 1 port_max_addrs 20 vlan_addr_ct 1: vlan_addr_max 20 total_addrs 4: max_total_addrs 6144

Port-security violation happened because MAC address has not been deleted from original port yet, hence "duplicate mac-address" message. To mitigate, but not completely alleviate the problem, we can reduce aging timer to 1 minute minimum. It still means that in case of fail-over, the floating IP address will not be accessible for another minute, which could be 1 minute too long.

Nexus: peer-switch and STP Bridge ID

With release of NX-OS 5.2, Cisco started supporting peer-switch feature on Nexus 5K. When peer-switch is enabled, both VPC primary and secondary switches originate STP BPDUs on vPC ports and use the same designated bridge ID on vPC ports. This got me wandering what brige ID vPC primary switch uses when peer-switch is not enabled. I set up vPC switch-pair with downstream switch connected via vPC port-channel. The switches are running MST. Here is partial BPDU captured on downstream Nexus switch with command:
ethanalyzer local interface inbound-hi display-filter "stp" limit-captured-frames 20

Spanning Tree Protocol
    Protocol Identifier: Spanning Tree Protocol (0x0000)
    Protocol Version Identifier: Multiple Spanning Tree (3)
    BPDU Type: Rapid/Multiple Spanning Tree (0x02)
    BPDU flags: 0x7c (Agreement, Forwarding, Learning, Port Role: Designated)
    Root Identifier: 8192 / 0 / 54:7f:ee:01:15:81
    Root Path Cost: 0
    Bridge Identifier: 8192 / 0 / 54:7f:ee:01:15:81
    Port identifier: 0x9063
    Message Age: 0
    Max Age: 20
    Hello Time: 2
    Forward Delay: 15
    Version 1 Length: 0
    Version 3 Length: 96
    MST Extension
        MST Config ID format selector: 0
        MST Config name: blp-mst-Region-1
        MST Config revision: 2
        MST Config digest: d7e7e4984e26acd301b955c5289031ad
        CIST Internal Root Path Cost: 0
        CIST Bridge Identifier: 8192 / 0 / 00:23:04:ee:be:01
            CIST Bridge Priority: 8192
            CIST Bridge Identifier System ID Extension: 0
            CIST Bridge Identifier System ID: 00:23:04:ee:be:01
        CIST Remaining hops: 20
        MSTID 1, Regional Root Identifier 8192 / 54:7f:ee:01:15:81
        MSTID 2, Regional Root Identifier 8192 / 54:7f:ee:01:15:81

Note "Bridge Identifier" and "CIST Bridge Identifier". They are different. The former is "vPC local system-mac" and latter is "vPC system-mac". They can be found in "show vpc role" output:

nexus-primary# show vpc role

vPC Role status
----------------------------------------------------
vPC role                        : primary                      
Dual Active Detection Status    : 0
vPC system-mac                  : 00:23:04:ee:be:01            
vPC system-priority             : 32667
vPC local system-mac            : 54:7f:ee:01:15:81            
vPC local role-priority         : 8192

Here we can see, that without peer-switch enabled Nexus switch uses 2 different bridge IDs in the same BPDU. Why does it do it? I reached out to Cisco and will update when I hear anything.
When peer-switch is enabled, both vPC primary and secondary switches originate BPDUs on vPC ports and "Bridge Identifier" and "CIST Bridge Identifier" are the same and equal to "vPC system-mac"