Category: networks

I have been cleaning up my email box and found some interesting stuff. This is from APNIC regarding a new approach to deploy multicast. Slides from nanog page (check tuesday) In my former employer, we suffered traffic congestion several times after some famous games got new updates. So it is interesting that Akamai is trying to deploy inter-domain multicast in the internet. They have a massive network and I guess they suffered with those updates and this is an attempt to “digest” better those spikes. At minute 16 you can see the network changes required. It doesnt look like a quick/easy change but would be a great thing to happen.

And reading a nanog conversation about 5G I realised that this technology promises high bandwidth (and that could fix the issue of requiring multicast). But still we should have a smarter way to deliver same content to eyeball networks?

From the nanog thread, there are several links to videos about 5G like this from verizon that gives the vision from a big provider and its providers (not technical). This one is more technical with 5G terms (I lost contact of Telco term with early 4G). As well, I see mentioning kubernetes in 5G deployments quite often. I guess something new to learn.

A good refresh about traceroute. It is a very common tool for network troubleshooting so it is important to use it wisely

Important points

• ICMP vs UDP: most implementations do UDP (it can be blocked…)
• Every probe is an independent trial!
• Try to identify the characteristics and location of each hop
• If there is a congestion/delay issue in one hop, it has to be carried out to the next hops, if not, it is just prioritizing of the ICMP generation by that router/hop.
• You dont see the reverse path – Ask the other end (if possible) to send the traceroute from its end.
• Border routers between providers can be a hot spot for issues.
• Asymmetric paths can bite you. Try to set the source address in your tests (from the provider IP, from your own space, etc)
• Spot ECMP (in the same hop, you see several different IPs). Multiple unequal length paths can be painful.
• MPLS: most times is hidden (TTL is removed). It can be tricky to spot. But it can be funny when you see the hops (with private IPs 🙂

And if you are more interested in the paths than latency, this can be a good too:

https://github.com/rucarrol/traceflow

I have struggled to get something working for learning a bit of openconfig.

At the end, all info my info comes from Anton’s Karneliuk blog series about Openconfig. So all credit to him. It is the best source about real testing of openconfig in different platforms.

OpenConfig. Part 1. OpenConfig (w/o and w/ Ansible) for Arista EOS, Cisco IOS XR and Nokia SR OS

I am not going to create the wheel and explain what openconfig is. In my head, it is attempt from several big vendors to standardise the network management (config) and monitoring (telemetry) via YANG (vendor-neutral) models. So OC uses YANG. And we interact with OC using a transport protocol like netconf, restcong and gNMI. So the network devices need to implement one of these protocols. Based on the blog Cisco, Nokia and Arista have netconf implementations and Ansible has a module for that!!! So the key words are openconfig, yang and netcong.

So in my case, based on my ceos lab, I have added a new playbook based on Anton’s to test openconfig/netcong with Arista cEOS:

https://github.com/thomarite/ceos-testing/blob/master/README.md#openconfig

This is quite basic as it only gets the interface config.

Following Anton’s Part3 Blog:

OpenConfig. Part 3. Advanced OpenConfig (w/ Ansible) for Arista EOS, Cisco IOS XR and Nokia SR OS: route policy BGP and interfaces again

I tried to push config via openconfig to my ceos devices (all files are in github as per link above).

The blog is dense but it is good because there is a lot of info. In this case, you have to use an Ansible role so it is a new thing to learn. As well, I wanted to adapt that role to my env and found some ansible issues but managed to fix after reading ansible documentation and paying attention to the -vvv info.

From “oc-push-config.yaml” the first task “collect” it is fine. It just takes some 10 minutes or more to get all YANG modules from each device.

The issue is with task “configure”. It fails when trying to push the interface config. I have tried Anton’s config and the actual config generated from oc-get-interface-info.yaml but no joy.

Based on the blog, it seems Arista doesnt have much interest in Openconfig.

Anyway, there have been a couple of intense days looking at all this openconfig/netcong/yang thing. I have just touched the surface but I have learned some Ansible in the way too. So could be worse.

Will move on to something else.

Another thing I wanted to play with is Netbox and found a good article to follow. So credits to the authors.

Using my current ceos lab from https://github.com/thomarite/ceos-testing

I followed Rick’s article to install netbox-docker and his own repo with the nornir examples using netbox. In this case nornir is going to use netbox as inventory. Normally I use local files. I created a python venv for 3.7.3

mkdir netbox-example; cd netbox-example
pyenv local 3.7.3
python -m virtualenv venv
source venv/bin/activate
git clone https://github.com/netbox-community/netbox-docker.git
cd netbox-docker
vim docker-compose.yml  --> so it always expose 8080
      nginx:
      ...
        ports:
          - 8080:8080
docker-compose pull
docker-compose up

When installing the requirements for “nornir-napalm-netbox-demo” I had to modify the version of some packages. So I removed the required version and I left pip to install the latest. I didnt use the makefile.

git clone https://github.com/rickdonato/nornir-napalm-netbox-demo
cd nornir-napalm-netbox-demo
python -m pip install -r requirements.txt

I struggled quite a bit with the management IP in netbox and the meaning of “platform”

• Create Manufacturers under Device Types: I created “Arista”
• Create Device Types under Device Types: I created “ceos”
• Create Platforms under Devices: This is VERY important as it has to be a supported NAPALM platform!!! So for Arista, I need “eos”.
• Create Device Roles under Devices. I created “pe”
• Create Devices under Devices.
• Within each device: add a management interface. Here, I got confused as I was adding the interface in the inventory section. The inventory section is info to/from the device using NAPALM. So you need to go to the bottom of the page, add the interface

• and then add an IP to that interface and mark it as primary.

Keep in mind that initially, I was using “0.0.0.0” for each device as that’s the IP I have be using for all my scripts lately.

Keep in mind (II) that we are using docker twice (from different commands…) one to get netbox and the other via docker(-topo) to get the Arista ceos containers…. and we have iptables rules under the hood created by both…

But, let’s go step by step. Now we need to confirm that our nornir scrip can connect to netbox. So follow “Nornir-to-Netbox Configuration” section. This is my file. I updated the nb_url and nb_token. Notice the usage of “transform_function“.

---
core:
num_workers: 20
inventory:
plugin: nornir.plugins.inventory.netbox.NBInventory
options:
nb_url: 'http://0.0.0.0:8080'
nb_token: '<NETBOX_API_TOKEN>'
ssl_verify: False
transform_function: "helpers.adapt_user_password"

You need to update “scripts/secrets.py” with the devices you have in your inventory and the user/pass:

creds = {
"r1": {"username": "user", "password": "pas123"},
"r2": {"username": "user", "password": "pas123"},
"r3": {"username": "user", "password": "pas123"},
}

So now we can test if nornir can connect to netbox:

/netbox-example/nornir-napalm-netbox-demo master$ python scripts/helpers.py --inventory
{'defaults': {'connection_options': {},
'data': {},
'hostname': None,
'password': None,
'platform': None,
'port': None,
'username': None},
'groups': {},
'hosts': {'r1': {'connection_options': {},
'data': {'asset_tag': 'r1',
'model': 'ceos',
'role': 'pe',
'serial': 'r1',
'site': 'lab1',
'vendor': 'Arista'},
'groups': [],
'hostname': '192.168.16.2',
'password': 'pas123',
'platform': 'eos',
'port': None,
'username': 'user'},
'r2': {'connection_options': {},
'data': {'asset_tag': 'r2',
'model': 'ceos',
'role': 'pe',
'serial': 'r2',
'site': 'lab1',
'vendor': 'Arista'},
'groups': [],
'hostname': '192.168.16.3',
'password': 'pas123',
'platform': 'eos',
'port': None,
'username': 'user'},
'r3': {'connection_options': {},
'data': {'asset_tag': 'r3',
'model': 'ceos',
'role': 'pe',
'serial': 'r3',
'site': 'lab1',
'vendor': 'Arista'},
'groups': [],
'hostname': '192.168.16.4',
'password': 'pas123',
'platform': 'eos',
'port': None,
'username': 'user'}}}

All good. Let’s see if we can get backups from the devices.

netbox-example/nornir-napalm-netbox-demo master$ python scripts/backup_configs.py
Backup Device configurations**
r1 ** changed : True *
vvvv Backup Device configurations ** changed : False vvvvvvvvvvvvvvvvvvvvvvvvvvv INFO
---- napalm_get ** changed : False --------------------------------------------- INFO
---- write_file ** changed : True ---------------------------------------------- INFO
^^^^ END Backup Device configurations ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
r2 ** changed : True *
vvvv Backup Device configurations ** changed : False vvvvvvvvvvvvvvvvvvvvvvvvvvv INFO
---- napalm_get ** changed : False --------------------------------------------- INFO
---- write_file ** changed : True ---------------------------------------------- INFO
^^^^ END Backup Device configurations ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
r3 ** changed : True *
vvvv Backup Device configurations ** changed : False vvvvvvvvvvvvvvvvvvvvvvvvvvv INFO
---- napalm_get ** changed : False --------------------------------------------- INFO
---- write_file ** changed : True ---------------------------------------------- INFO
^^^^ END Backup Device configurations ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(venv) /netbox-example/nornir-napalm-netbox-demo master$

Great all good.

Now, let’s see netbox using NAPALM. If you click on “Status” for any device, netbox will use NAPALM to get the facts from the device. If netbox is not configured properly with NAPALM, it will fail. This is a working scenario:

The tabs “LLDP neighbors” and “Configuration” relay too in NAPALM.

So for configuring netbox with napalm you need to tell netbox the user/pass that NAPALM needs:

netbox-example$ vim netbox-docker/env/netbox.env
...
NAPALM_USERNAME=user
NAPALM_PASSWORD=pas123
NAPALM_TIMEOUT=10
...

Very likely you will have to restart netbox:

/netbox-docker release$ docker-compose down
/netbox-docker release$ docker-compose up

As mentioned before, I had an issue when I was using “0.0.0.0” as IP. By default (as It seems I can’t think) I was using the exposed IP/port from docker-topo to reach the ceos switches. I haven’t had an issue until using netbox.

I am using docker for netbox and docker(-topo) for my arista cEOS switches. So the connectivity between netbox and ceos is via the IPs/interfaces/bridges created by docker. And remember… you have iptables under the hood. My first mistake was telling netbox to use 0.0.0.0 as it is the one I am using to testing from my scripts when connecting to ceos. Netbox needs to point to the IP assigned by docker :facepalm: 192.16.16.x in my case. Second one, the port, same thing docker exposes the port 443 as 900x for external connections and I use 900x in my scripts. From netbox point of view, it is still 443 :facepalm: And finally, I am calling docker twice for building my lab, one for netbox-docker and the other for ceos. You need to keep an eye on iptables changes when restarting netbox via docker-compose because you can be in the situation that netbox traffic is dropped in DOCKER-ISOLATION-STAGE-1 :facepalm: (need to try to write a docker-compose to build everything in one go)

So when I was having errors from netbox that it was being rejected when connecting to ceos devices via NAPALM, I couldnt understand it. My scripts were fine using those details (0.0.0.0:900x)

I ran tcpdump on one ceos on the “ethernet0” interface and NOTHING was hitting the interface from netbox on port 900x but my scripts could…..

Somehow netbox wasnt able to reach ceos r1??? In my head, netbox and ceos devices were all in 0.0.0.0….. so no routing, no firewalls, they are connected in the same network 0.0.0.0…..

At the end I waked up and realised that the docker devices are using the IPs provided by docker so it is following normal routing… and firewalling by iptables. The same for ceos devices, they have IPs (different from 0.0.0.0)

So I updated netbox with the correct management IPs for r1, r2 and r3 ceos.

When I filtered by the real netbox IP in r1 tcpdump ethernet0, I was seeing traffic on 900x!!! Good. Then I realised that it has to be 443. So I removed my hack to update the port to 900x.

For a different reason I had to restart docker-topo (for ceos) and then docker netbox. And now, I coudnt see any traffic from netbox in r1….. I “didn’t” change anything. So the routing didnt change, there was something else “cutting” the connection: iptables

docker uses iptables very heavily. I realised that after restart docker-netbox, iptables changed…

before restart:

# iptables -t filter -S DOCKER-ISOLATION-STAGE-1
Warning: iptables-legacy tables present, use iptables-legacy to see them
-N DOCKER-ISOLATION-STAGE-1
-A DOCKER-ISOLATION-STAGE-1 -j ACCEPT
-A DOCKER-ISOLATION-STAGE-1 -i br-94a8183a4fb1 ! -o br-94a8183a4fb1 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-0d4ec9aba9bd ! -o br-0d4ec9aba9bd -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-609619313dc8 ! -o br-609619313dc8 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-61d32350cb58 ! -o br-61d32350cb58 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-384488acbc99 ! -o br-384488acbc99 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i docker0 ! -o docker0 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -j RETURN

after restart:

# iptables -t filter -S DOCKER-ISOLATION-STAGE-1
Warning: iptables-legacy tables present, use iptables-legacy to see them
-N DOCKER-ISOLATION-STAGE-1
-A DOCKER-ISOLATION-STAGE-1 -i br-381cdff63d2f ! -o br-381cdff63d2f -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -j ACCEPT
-A DOCKER-ISOLATION-STAGE-1 -i br-94a8183a4fb1 ! -o br-94a8183a4fb1 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-0d4ec9aba9bd ! -o br-0d4ec9aba9bd -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-609619313dc8 ! -o br-609619313dc8 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-61d32350cb58 ! -o br-61d32350cb58 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i docker0 ! -o docker0 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -j RETURN

With the restart, docker created a new bridge interface for netbox (old: br-384488acbc99, new: br-381cdff63d2f) and it wasnt hitting anymore the “DOCKER-ISOLATION-STAGE-1 -j ACCEPT”

So I had to make an iptables change:

# iptables -t filter -D DOCKER-ISOLATION-STAGE-1 -j ACCEPT
# iptables -t filter -I DOCKER-ISOLATION-STAGE-1 -j ACCEPT
# iptables -t filter -S DOCKER-ISOLATION-STAGE-1
Warning: iptables-legacy tables present, use iptables-legacy to see them
-N DOCKER-ISOLATION-STAGE-1
-A DOCKER-ISOLATION-STAGE-1 -j ACCEPT
-A DOCKER-ISOLATION-STAGE-1 -i br-381cdff63d2f ! -o br-381cdff63d2f -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-94a8183a4fb1 ! -o br-94a8183a4fb1 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-0d4ec9aba9bd ! -o br-0d4ec9aba9bd -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-609619313dc8 ! -o br-609619313dc8 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i br-61d32350cb58 ! -o br-61d32350cb58 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -i docker0 ! -o docker0 -j DOCKER-ISOLATION-STAGE-2
-A DOCKER-ISOLATION-STAGE-1 -j RETURN
#

And finally, netbox could use napalm to contact the ceos devices…. Calling docker twice is not a great idea….

BTW, this is my docker ps with netbox and ceos devices:

(venv) /netbox-example$ docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
c23be76ffd54 nginx:1.17-alpine "nginx -c /etc/netbo…" 4 hours ago Up 4 hours 80/tcp, 0.0.0.0:8080->8080/tcp netbox-docker_nginx_1
5a0b89f18578 netboxcommunity/netbox:latest "/opt/netbox/docker-…" 4 hours ago Up 4 hours netbox-docker_netbox_1
528948de329b netboxcommunity/netbox:latest "python3 /opt/netbox…" 4 hours ago Up 4 hours netbox-docker_netbox-worker_1
29529302ba1c redis:5-alpine "docker-entrypoint.s…" 4 hours ago Up 4 hours 6379/tcp netbox-docker_redis_1
5e975ec2aa70 redis:5-alpine "docker-entrypoint.s…" 4 hours ago Up 4 hours 6379/tcp netbox-docker_redis-cache_1
6158672a4ae6 postgres:11-alpine "docker-entrypoint.s…" 4 hours ago Up 4 hours 5432/tcp netbox-docker_postgres_1
34841aa098d4 ceos-lab:4.23.3M "/sbin/init systemd.…" 5 hours ago Up 5 hours 0.0.0.0:2002->22/tcp, 0.0.0.0:9002->443/tcp 3node_r03
4ca92c6a3b09 ceos-lab:4.23.3M "/sbin/init systemd.…" 5 hours ago Up 5 hours 0.0.0.0:2001->22/tcp, 0.0.0.0:9001->443/tcp 3node_r02
67e8b7ab84e0 ceos-lab:4.23.3M "/sbin/init systemd.…" 5 hours ago Up 5 hours 0.0.0.0:2000->22/tcp, 0.0.0.0:9000->443/tcp 3node_r01

I was painful but I learned a couple of things about netbox, nornir and docker/iptables!!!

Reflect:
client -> clients and non-clients
non-client -> clients
No Reflect:
non-client -> no-client (normal ibgp)

always advertises to eBGP peers (normal ebgp)
eBGP learned prefixes, advertised to client and non-clients (normal ebgp)

Full Mesh iBGP:

• between RRs
• RRs and non-clients
• clients just need iBGP to RRs.

Reflects ONLY the best route

RRs dont modify on reflected routes: NEXT_HOP, AS_PATH, LP and MED.

Prevent routing information loops: ORIGINATOR_ID and CLUSTER_LIST

Clustering:

• ORIGINATOR_ID: The first RR creates the Originator_ID and sets it to the BGP router ID of the router that originates the route. So when a client receives a route with its own Originator ID, it is dropped.
• CLUSTER_LIST: if the local CLUSTER_ID is found in the list, the route is discarded.
  *This is done ONLY in RRs.
  This is ONLY created or updated on a RR during Reflection.

Hierarchical Route Reflection:
2 levels:
- level1 RRs are clients of level2 RR -> level1 RR dont need full mesh between them
- level2 RRs are full mesh between them.

Consider:
size of top-level mesh
number of alternative paths

Hierarchical Route Reflection:
2 levels:

– level1 RRs are clients of level2 RR -> level1 RR dont need full mesh between them
– level2 RRs are full mesh between them.

Consider:

• size of top-level mesh
• number of alternative paths

RR Design Priciples:

• keep logical and physical topologies congruent to increase redundancy and path optimization and prevent loops
  — follow physical topology
  — change physical topology
  — modify logical topology
  — follow physical topology
  — session between RR and non-client shouldnt traverse a client
  — session between RR and client shouldnt traverse a non-client

• use comparable metrics in route selection to avoid convergence oscillations
  (ie MED – only used between prefixes from same neighbor AS – not used for different AS)
  — full ibgp mess –> no
  — always-compare-med –> no
  — deterministic-med -> ok
  — med=0 (via RM in) -> ok
  — bgp communities ->

• set proper intra and inter cluster IGP metrics to avoid convergence oscillations
  — multicluster RR architecture: intracluster metrics lower than intercluster.

• modify next-hop with care. Do so only to bring the RRs into forwarding path.

• use peer groups with RR to reduce convergence time.