LLM n C, 1.6nm, xz vul, turing 2024, let’s encrypt, chatdev, Ethernet vs IB, Slingshot, Tailscale ssh, videos, 42 rules, CNI, Cilium

Origins of deep learning: interesting post. At the beginning all was Matlab and CPU bounded. repo

LLM in C: post and repo.

A16: 1.6nm process for 2026. More frequency, less power.

xz vulnerability repo: Something I need to check in the VP

Turing Award 2024: zero-knowledge-proof.

Cloudflare and Let’s Encrypt’s certificate change: I haven’t heard of this until recently. I use Let’s Encrypt so as far as I can read, makes sense what they are doing. But didnt know 2% Cloudflare customer were using the “cert”

ChatDev: Communicate agents for software development. I am a not a developer but I would use this just as a starting point If I have any idea for a project. I would remove the C-suite agents, at least for low level projects.

IB vs Ethernet: A bit of bias here (the author is from Broadcom -> Ethernet). I have no hands-on experience with IB, but I have read the cables are not cheap… Let’s see when UltraEthernet gets into the market. Another view.

Slingshot and Juniper: A bit of bias again as HP bought Juniper. So how will these interconnects fade inside the company? As far as I know, most supercomputers use some “special” interconnect so not much ethernet there. But the money nowadays is in AI infra… Paper for slingshot (haven’t read it)

Tailscale SSH, wireguard throughput: These are things I should a spend a bit of time one day and consider if I should use them (I dont like it is not opensource though). This netmaker?


Jocko Willink: Discipline = Freedom. Remember but not dwell. Good leader, delegate. Be a man -> take action, bonding (pick your activity)

Jimmy Carr: Imposter syndrome each 18 months, so you have to stand-up. People crave the success not the journey. Teaching comedy good for communicating.

Sam Altman – Stanford 2024: First time I see him talking. It has some funny moments. More powerful computers. I missed a question about opensource LLM and closed ones.

Find a girlfriend: I know just a little bit about the person (I want to read one of his books) from other books and videos. I would think he would have already a girlfriend or family. From the three methods, definitely, the face to face approach in the street looks so much better (and that’s what I would like to do)

Jordan Peterson original 42 rules

CNI performance: I have used kubernetes since I studied for CKAD but still I am interested in the networks side. I didn’t know about Kube-router and it did great! I am bit surprised with Calico as I have read more and more about Cilium.

Cilium for network engineers. I have to read this fully (worried that Cisco bought it…)

Life, Love, Sex, Negative Beliefs, startup regrets, nanog90, Groq LPU, LLM from scratch, ssh3, eBFP BGP, RPKI, TIANHE-3

I hit rock bottom this week. I hope I finally closed one door in my life so I give myself the chance to open others. Made the wrong decision? It is easy when you look back. Do I regret it? The most annoying thing is these are failures so you can’t go back and recover. But I was so bloody newbie!!!…. At least after 5 years…

“For every reason it’s not possible, there are hundreds of people who have faced the same circumstances and succeeded.” Jack Canfield

Head down, crying, cursing, whatever, but forwards. As it has always been.


Somehow managed to list to long videos, something I normally can’t manage (because lack of time, etc)

Negative Beliefs, avoid bitterness, aim for greatness (remarkable things), scape the darkness: Jordan B Peterson with Modern Wisdom: video, podcast.

Find and keep Love: video. 1st Get your shit together. Communication is critical. Be careful with your shopping list….

Good Sex: video. Communicate….

Orgasm: video. Haven’t seen it completely yet but very interesting. Use your tongue wisely.

— Other things:

Startup decisions and regrets: page. Interesting. I think most of things are very specific but still good to read.

Nanog90: agenda I didnt want the videos but I reviewed several pdfs and these ones look interesting:

Abstract Ponderings: A ten-year retrospective. Rob Shakir – Google: video


AI Data Center networks – Juniper – video

Using gNOI capabilities to simplify software upgrade use case: video – I had to idea about gNOI so looks interesting. It is crazy that still in XXI, automating a network device is so painful. Thanks to all vendors to make your life miserable.

Go lang for network engineers: video slides– I always thought that Golang had a massive potential for network automation but there was always lack of support and python is the king. So nice to see that Arista has things to offer.

PTP in Meta: video and blog.

There are more things, but havent had the chance to review them.


It looks there is new chatbot that is not using the standard NVIDIA GPU. Groq uses LPU (Language Processing Unit). And they say it is better than a GPU. They have this paper but I can’t really see feature of that LPU.

Slurp’it: Show this blog, and the product looks interesting but although is free, it is not opensource and at the end of they you dont want a new vendor-lockin

Container lab in kubernetes: Clabernetes. I would like to play with this one day.

NetDev0x17: videos and sessions. link This is quite low details and most of the time beyond my knowledge. Again, something to take a look at some point.

LLM from scratch: repo. Looks very interesting. But the book it is going to take a long time to hit the market.

ssh3: repo. Interesting experiment.

eBFP and BGP: blog. Really interesting. Another thing that always wanted to play with.

Orange RPKI: old news but still interesting to see how much damaged can cause RPKI in the wrong hands…

China TIANHE-3 Supercomputer: Very interesting. Link.

OpenAI / ChatGPT

I guess this is a bit old news already. But I wanted to test it for some weeks. I have just made one question so far (and it failed a couple of times until I got an answer.

And my very first question was to list the most famous quotes from Aristotle to his actual writings. Why? Because you read many quotes from many famous people that are not actually true and it is very hard to really track if you dont read all books from that person. And to be honest, philosophy is not easy to digest.

Here are a few famous quotes from Aristotle that reference his actual writings:

"Happiness is the meaning and the purpose of life, the whole aim and end of human existence." - Nicomachean Ethics

"It is the mark of an educated mind to be able to entertain a thought without accepting it." - Metaphysics

"We are what we repeatedly do. Excellence, then, is not an act, but a habit." - Nicomachean Ethics

"The whole is greater than the sum of its parts." - Metaphysics

"Nature does nothing in vain." - Politics

Please note that these quotes should not be taken as direct verbatim from Aristotle's work but rather as a representation of his thoughts.

Apart from that, I wonder how is the infrastructure of this company? Is on premise or cloud? How is their DC network design? TCP? Checking one of the open positions it seems they have a pretty big cluster. Although they mention cloud platforms. And the company has some kind of agreement with MS (so Azure is going to be used)

IPv6 BIG TCP / Replace TCP in DC: Homa

This week a colleague pass this link about running kubernetes cluster running on Cilium. The interesting point is the high throughput is achieved by BIG TCP and IPv6!

The summary (copied) is:

TCP segments in the OS are up to 65K, NIC hardware does the segmentation – we do this now, but the 65K is a limitation of IPv4 addressing.  BIG TCP uses IPv6 and allows much large TCP segments within OS currently 512K but theoretically higher.  End result – better perf (>20% higher in this video) and latency (2.2x faster through the OS).

Then I saw this other video from John Ousterhout. It is similar topic as the Kubernetes video above as K8S is used mainly in datacenters.

High performance:
– data throughput: full link speed for large messages
– low tail latency: <10us for short messages? (DC)
– message throughput: 100M short messages per second? (DC)

TCP issues in DC:
1- stream oriented (no load balancing) -> message based
2- connection oriented (can break infiniband!, expensive,)-> connectionless
3- fair scheduling (bw sharing) -> run to completion (SRPT)
4- sender-driven congestion control (based on buffer occupancy) -> receiver- driven congestion control
5- in-order delivery -> no ordering requirements

As well, it is important the move to NIC (as there is already a lot of NIC offloading).

His proposal for HOMA looks very nice but I like how he explains how dificult is going to be successful. Still worth trying.

ARP Storms – EVPN

We have had an issue with broadcast storms in our network. Checking the CoPP setup in the switches, we could see massive drops of ARP. This is a good link to know how to check CoPP drops in NXOS.

N9K:# show copp status
N9K# show policy-map interface control-plane | grep 'dropped [1-9]' | diff

Having so many ARP drops by CoPP is bad because very likely good ARP requests are going to be dropped.

Initially i thought it was related to ARP problems in EVPN like this link. But after taking a packet capture in a switch from an interface connected to a server, I could see that over 90% ARP traffic coming from the server was not getting a reply…. Checking in different switches, I could see the same pattern all over the place.

So why the server was making so many ARP requests?

After some time, managed to help help from a sysadmin with access to the servers so could troubleshoot the problem.

But, how do you find the process that is triggering the ARP requests? I didnt make the effort to think about it and started to search for an easy answer. This post gave me a clue.

ss does show you connections that have not yet been resolved by arp. They are in state SYN-SENT. The problem is that such a state is only held for a few seconds then the connection fails, so you may not see it. You could try rapid polling for it with

while ! ss -p state syn-sent | grep; do sleep .1; done

Somehow I couldnt see anything anything with “ss” so tried netstat as it shows you too the status of the TCP connection (I wonder what would happen is the connection was UDP instead???)

Initially I tried “netstat -a” and it was too slow to show me “SYN-SENT” status

Shame on me, I had to search how to get to show the ports quickly here:

watch netstat -ntup | grep -i syn_sent | awk '{print $4,$5,$6,$7}'

It was slow because it was trying to resolve all IPs to hostname…. :facepalm. Tha is fixed with “-n” (no-resolve)

Anyway, with the command above, finally managed to see the process that were in “SYN_SENT” state

This is not the real thing, just an example:

#  netstat -ntup | grep -i syn_sent 
tcp        0      1              SYN_SENT    98690/telnet        

We could see that the destination port was TCP 179, so something in the node was trying to talk BGP! They were “bird” processes. As the node belonged to a kubernetes cluster, we could see a calico container as CNI. Then we connected to the container and tried to check the bird config. We could see clearly the IPs that dont get ARP reply were configured there.

So in summary, basic TCP:

Very summarize, TCP is L4, then goes down to L3 IP. For getting to L2, you need to know the MAC of the IP, so that triggers the ARP request. Once the MAC is learned, it is cached for the next request. For that reason the first time you make a connection is slow (ping, traceroute, etc)

Now we need to workout why the calico/bird config is that way. Fix it to only use IPs of real BGP speakers and then verify the ARP storms stop.

Hopefully, I will learn a bit about calico.

Notes for UDP:

If I generate an UDP connection to a non-existing IP

$ nc -u 4000

netstat tells me the UDP connection is established and I can’t see anything in the ARP table for an external IP, for an internal IP (in my own network) I can see an incomplete entry. Why?

#  netstat -ntup | grep -i
udp        0      0            ESTABLISHED 102014/nc           
#  netstat -ntup | grep -i ''
udp        0      0        ESTABLISHED 102369/nc           
# arp -a
? ( at <incomplete> on wlp2s0
something.mynet ( at xx:xx:xx:yy:yy:zz [ether] on wlp2s0

# tcpdump -i wlp2s0 host
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on wlp2s0, link-type EN10MB (Ethernet), snapshot length 262144 bytes
23:35:45.081819 IP > UDP, length 1
23:35:45.081850 IP > UDP, length 1
23:35:46.082075 IP > UDP, length 1
23:35:47.082294 IP > UDP, length 1
23:35:48.082504 IP > UDP, length 1
5 packets captured
5 packets received by filter
0 packets dropped by kernel
  • UDP is stateless so we can’t have states…. so it is always going to be “established”. Basic TCP/UDP
  • When trying to open an UDP connection to an external IP, you need to “route” so my laptop knows it needs to send the UDP connection to the default gateway, so when getting to L2, the destination MAC address is not is the default gateway MAC. BASIC ROUTING !!!! For that reason you dont see in ARP table
    • When trying to open an UDP connection to a local IP, my laptop knows it is in the same network so it should be able to find the destination MAC address using ARP.


This week I read that kubernetes is going to stop support for Docker soon. I was quite surprised. I am not an expert so it seems they have legit reasons. But I haven’t read anything from the other side. I think it is going to be painful so I need to try that in my lab and see how to do that migration. It has to be nice to learn that.

In the other end, I read a blog entry about ASICs from Cloudflare. I think without getting too technical it is a good one. And I learn about the different type of ASICs from Juniper. In the last years, I have only used devices powered by Broadcom ASICs. One day, I would like to try that P4/Barefoot Tofino devices. And related to this, I remember this NANOG presentation about ASICs that is really good (and fun!).


While studying for CKA, I installed kubeadm using vagrant/virtualbox. Now I want to try the same, but using libvirt instead.

1- Install 3VM (1 master and 2 worker-nodes) I have installed vagrant and libvirtd already. Take this vagrant file as source.

2- I had to make two changes to that file

2.1- I want to use libvirtd, so need to change the Ubuntu vm.box to one that supports it.

#config.vm.box = “ubuntu/bionic64”
config.vm.box = “generic/ubuntu1804”

2.2- Then need to change the network interface

enp0s8 -> eth1

3- Create the VMs with vagrant.

$ ls -ltr
-rw-r--r-- 1 tomas tomas 3612 Nov 15 16:36 Vagrantfile

$ vagrant status
Current machine states:
kubemaster not created (libvirt)
kubenode01 not created (libvirt)
kubenode02 not created (libvirt)

$ vagrant up
An unexpected error occurred when executing the action on the
'kubenode01' machine. Please report this as a bug:
cannot load such file -- erubis

3.1 Ok, we have to troubleshoot vagrant in my laptop. I googled a bit and couldnt find anything related. I remembered that you could install plugins with vagrant as once I had to update vagrant-libvirtd plugin. So this is kind of what I did.

$ vagrant version
Installed Version: 2.2.13
Latest Version: 2.2.13

$ vagrant plugin list
vagrant-libvirt (0.1.2, global)
Version Constraint: > 0

$ vagrant plugin update
Updating installed plugins…
Fetching fog-core-2.2.3.gem
Fetching nokogiri-1.10.10.gem
Building native extensions. This could take a while…
Building native extensions. This could take a while…
Fetching vagrant-libvirt-0.2.1.gem
Successfully uninstalled excon-0.75.0
Successfully uninstalled fog-core-2.2.0
Removing nokogiri
Successfully uninstalled nokogiri-1.10.9
Successfully uninstalled vagrant-libvirt-0.1.2
Updated 'vagrant-libvirt' to version '0.2.1'!

$ vagrant plugin install erubis

$ vagrant plugin update
Updating installed plugins…
Building native extensions. This could take a while…
Building native extensions. This could take a while…
Updated 'vagrant-libvirt' to version '0.2.1'!

$ vagrant plugin list
erubis (2.7.0, global)
Version Constraint: > 0
vagrant-libvirt (0.2.1, global)
Version Constraint: > 0

3.2. Now, I can start vagrant fine

$ vagrant up

$ vagrant status
Current machine states:
kubemaster running (libvirt)
kubenode01 running (libvirt)
kubenode02 running (libvirt)

4- Install kubeadm. I follow the official doc. It seems we have the pre-requisites. My laptop has 8GB RAM and 4 cpus. Our VMs are Ubuntu 16.04+.

4.1 Enable iptables in each VM:

$ vagrant ssh kubemaster

vagrant@kubemaster:~$ lsmod | grep br_net
vagrant@kubemaster:~$ sudo modprobe br_netfilter
vagrant@kubemaster:~$ lsmod | grep br_net
br_netfilter 24576 0
bridge 155648 1 br_netfilter
vagrant@kubemaster:~$ cat <<EOF | sudo tee /etc/sysctl.d/k8s.conf
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
vagrant@kubemaster:~$ sudo sysctl --system

5- Install runtime (docker). Following the official doc, we click on the link at the end of “Installing runtime”. We do this in each node:

vagrant@kubemaster:~$ sudo -i
root@kubemaster:~# sudo apt-get update && sudo apt-get install -y apt-transport-https ca-certificates curl software-properties-common
root@kubemaster:~# curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key --keyring /etc/apt/trusted.gpg.d/docker.gpg add -
root@kubemaster:~# sudo add-apt-repository \
"deb [arch=amd64] https://download.docker.com/linux/ubuntu \ 
$(lsb_release -cs) \
root@kubemaster:~# sudo apt-get update && sudo apt-get install -y \
containerd.io=1.2.13-2 \
docker-ce=5:19.03.11~3-0~ubuntu-$(lsb_release -cs) \
docker-ce-cli=5:19.03.11~3-0~ubuntu-$(lsb_release -cs)
root@kubemaster:~# cat <<EOF | sudo tee /etc/docker/daemon.json
"exec-opts": ["native.cgroupdriver=systemd"],
"log-driver": "json-file",
"log-opts": {
"max-size": "100m"
"storage-driver": "overlay2"
"exec-opts": ["native.cgroupdriver=systemd"],
"log-driver": "json-file",
"log-opts": {
"max-size": "100m"
"storage-driver": "overlay2"
root@kubemaster:~# sudo mkdir -p /etc/systemd/system/docker.service.d
root@kubemaster:~# sudo systemctl daemon-reload
root@kubemaster:~# sudo systemctl restart docker
root@kubemaster:~# sudo systemctl enable docker
Synchronizing state of docker.service with SysV service script with /lib/systemd/systemd-sysv-install.
Executing: /lib/systemd/systemd-sysv-install enable docker

5- Now we follow “Installing kubeadm, kubelet and kubectl” from main doc in each VM.

root@kubemaster:~# sudo apt-get update && sudo apt-get install -y apt-transport-https curl
root@kubemaster:~# curl -s https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -
root@kubemaster:~# cat <<EOF | sudo tee /etc/apt/sources.list.d/kubernetes.list
deb https://apt.kubernetes.io/ kubernetes-xenial main
deb https://apt.kubernetes.io/ kubernetes-xenial main
root@kubemaster:~# sudo apt-get update
root@kubemaster:~# sudo apt-get install -y kubelet kubeadm kubectl
root@kubemaster:~# ip -4 a

We dont have to do anything with the next section “Configure cgroup driver…” as we are using docker. So from the bottom of the main page, we click on the next section for using kubeadm and create a cluster.

6- So we have our three VMS with kubeadm. Now we are going to create a cluster. The kubemaster VM will be the control-plane node. So following “Initializing your control-plane node”, we dont need 1 (as we have only one control-node), for 2) will install weave-net as CNI in the next step, we need to use a new network for this: 3) we dont need it and 4) we will specify the master ip. So, only on kubemaster:

root@kubemaster:~# kubeadm init --pod-network-cidr --apiserver-advertise-address=
W1115 17:13:31.213357 9958 configset.go:348] WARNING: kubeadm cannot validate component configs for API groups [kubelet.config.k8s.io kubeproxy.config.k8s.io]
[init] Using Kubernetes version: v1.19.4
[preflight] Running pre-flight checks
error execution phase preflight: [preflight] Some fatal errors occurred:
[ERROR Swap]: running with swap on is not supported. Please disable swap
[preflight] If you know what you are doing, you can make a check non-fatal with --ignore-preflight-errors=...
To see the stack trace of this error execute with --v=5 or higher

oh, problem. It seems we need to disable swap on the VMs. Actually, we will do in all VMs.

root@kubemaster:~# swapoff -a

Try again kubeadm init in master:

root@kubemaster:~# kubeadm init --pod-network-cidr --apiserver-advertise-address=
W1115 17:15:00.378279 10376 configset.go:348] WARNING: kubeadm cannot validate component configs for API groups [kubelet.config.k8s.io kubeproxy.config.k8s.io]
[init] Using Kubernetes version: v1.19.4
[preflight] Running pre-flight checks
[preflight] Pulling images required for setting up a Kubernetes cluster
[preflight] This might take a minute or two, depending on the speed of your internet connection
[preflight] You can also perform this action in beforehand using 'kubeadm config images pull'
[certs] Using certificateDir folder "/etc/kubernetes/pki"
[certs] Generating "ca" certificate and key
[certs] Generating "apiserver" certificate and key
[certs] apiserver serving cert is signed for DNS names [kubemaster kubernetes kubernetes.default kubernetes.default.svc kubernetes.default.svc.cluster.local] and IPs []
[certs] Generating "apiserver-kubelet-client" certificate and key
[certs] Generating "front-proxy-ca" certificate and key
[certs] Generating "front-proxy-client" certificate and key
[certs] Generating "etcd/ca" certificate and key
[certs] Generating "etcd/server" certificate and key
[certs] etcd/server serving cert is signed for DNS names [kubemaster localhost] and IPs [ ::1]
[certs] Generating "etcd/peer" certificate and key
[certs] etcd/peer serving cert is signed for DNS names [kubemaster localhost] and IPs [ ::1]
[certs] Generating "etcd/healthcheck-client" certificate and key
[certs] Generating "apiserver-etcd-client" certificate and key
[certs] Generating "sa" key and public key
[kubeconfig] Using kubeconfig folder "/etc/kubernetes"
[kubeconfig] Writing "admin.conf" kubeconfig file
[kubeconfig] Writing "kubelet.conf" kubeconfig file
[kubeconfig] Writing "controller-manager.conf" kubeconfig file
[kubeconfig] Writing "scheduler.conf" kubeconfig file
[kubelet-start] Writing kubelet environment file with flags to file "/var/lib/kubelet/kubeadm-flags.env"
[kubelet-start] Writing kubelet configuration to file "/var/lib/kubelet/config.yaml"
[kubelet-start] Starting the kubelet
[control-plane] Using manifest folder "/etc/kubernetes/manifests"
[control-plane] Creating static Pod manifest for "kube-apiserver"
[control-plane] Creating static Pod manifest for "kube-controller-manager"
[control-plane] Creating static Pod manifest for "kube-scheduler"
[etcd] Creating static Pod manifest for local etcd in "/etc/kubernetes/manifests"
[wait-control-plane] Waiting for the kubelet to boot up the control plane as static Pods from directory "/etc/kubernetes/manifests". This can take up to 4m0s
[apiclient] All control plane components are healthy after 25.543262 seconds
[upload-config] Storing the configuration used in ConfigMap "kubeadm-config" in the "kube-system" Namespace
[kubelet] Creating a ConfigMap "kubelet-config-1.19" in namespace kube-system with the configuration for the kubelets in the cluster
[upload-certs] Skipping phase. Please see --upload-certs
[mark-control-plane] Marking the node kubemaster as control-plane by adding the label "node-role.kubernetes.io/master=''"
[mark-control-plane] Marking the node kubemaster as control-plane by adding the taints [node-role.kubernetes.io/master:NoSchedule]
[bootstrap-token] Using token: aeseji.kovc0rjt6giakn1v
[bootstrap-token] Configuring bootstrap tokens, cluster-info ConfigMap, RBAC Roles
[bootstrap-token] configured RBAC rules to allow Node Bootstrap tokens to get nodes
[bootstrap-token] configured RBAC rules to allow Node Bootstrap tokens to post CSRs in order for nodes to get long term certificate credentials
[bootstrap-token] configured RBAC rules to allow the csrapprover controller automatically approve CSRs from a Node Bootstrap Token
[bootstrap-token] configured RBAC rules to allow certificate rotation for all node client certificates in the cluster
[bootstrap-token] Creating the "cluster-info" ConfigMap in the "kube-public" namespace
[kubelet-finalize] Updating "/etc/kubernetes/kubelet.conf" to point to a rotatable kubelet client certificate and key
[addons] Applied essential addon: CoreDNS
[addons] Applied essential addon: kube-proxy
Your Kubernetes control-plane has initialized successfully!
To start using your cluster, you need to run the following as a regular user:
mkdir -p $HOME/.kube
sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
sudo chown $(id -u):$(id -g) $HOME/.kube/config
You should now deploy a pod network to the cluster.
Run "kubectl apply -f [podnetwork].yaml" with one of the options listed at:
Then you can join any number of worker nodes by running the following on each as root:
kubeadm join --token aeseji.kovc0rjt6giakn1v \
--discovery-token-ca-cert-hash sha256:c1b91ec9cebe065665c314bfe9a7ce9c0ef970d56ae762dae5ce308caacbd8cd

7- We need to follow the output of kubeadm init in kubemaster. As well pay attention as the info for joining our worker-nodes to the cluster in there too (“kubeadm join ….”)

root@kubemaster:~# exit
vagrant@kubemaster:~$ mkdir -p $HOME/.kube
vagrant@kubemaster:~$ sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
vagrant@kubemaster:~$ sudo chown $(id -u):$(id -g) $HOME/.kube/config

We can test the status of the control-node. It is NotReady because it needs the network configuration.

vagrant@kubemaster:~$ kubectl get nodes
kubemaster NotReady master 2m9s v1.19.4

8- From the same page, now we need to follow “Installing a Pod network add-on”. I dont know why but the documentation is not great about it. You need to dig in all version to find the steps to install wave-net. This is the link. So we install wave-net only on the kubemaster:

vagrant@kubemaster:~$ kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')"
serviceaccount/weave-net created
clusterrole.rbac.authorization.k8s.io/weave-net created
clusterrolebinding.rbac.authorization.k8s.io/weave-net created
role.rbac.authorization.k8s.io/weave-net created
rolebinding.rbac.authorization.k8s.io/weave-net created
daemonset.apps/weave-net created
vagrant@kubemaster:~$ kubectl get nodes
kubemaster Ready master 4m32s v1.19.4

9- We can follow to the section “Joining your nodes”. We need to apply the “kubeadm join…” command from the outout of “kubeadm init” in master node in only the worker-nodes.

root@kubenode02:~# kubeadm join --token aeseji.kovc0rjt6giakn1v --discovery-token-ca-cert-hash sha256:c1b91ec9cebe065665c314bfe9a7ce9c0ef970d56ae762dae5ce308caacbd8cd
[preflight] Running pre-flight checks
[preflight] Reading configuration from the cluster…
[preflight] FYI: You can look at this config file with 'kubectl -n kube-system get cm kubeadm-config -oyaml'
[kubelet-start] Writing kubelet configuration to file "/var/lib/kubelet/config.yaml"
[kubelet-start] Writing kubelet environment file with flags to file "/var/lib/kubelet/kubeadm-flags.env"
[kubelet-start] Starting the kubelet
[kubelet-start] Waiting for the kubelet to perform the TLS Bootstrap…
This node has joined the cluster:
Certificate signing request was sent to apiserver and a response was received.
The Kubelet was informed of the new secure connection details.
Run 'kubectl get nodes' on the control-plane to see this node join the cluster.

10- We need to wait a bit, but finally the worker nodes will come up as Ready if we check in the master/control-node:

vagrant@kubemaster:~$ kubectl get nodes
kubemaster Ready master 6m35s v1.19.4
kubenode01 Ready 2m13s v1.19.4
kubenode02 Ready 2m10s v1.19.4

11- Let’s verify we have a working cluster just creating a pod.

vagrant@kubemaster:~$ kubectl run ngix --image=nginx
pod/ngix created

vagrant@kubemaster:~$ kubectl get pod
ngix 0/1 ContainerCreating 0 5s
vagrant@kubemaster:~$ kubectl get pod
ngix 1/1 Running 0 83s

vagrant@kubemaster:~$ kubectl delete pod ngix
pod "ngix" deleted

vagrant@kubemaster:~$ kubectl get pod -n kube-system
coredns-f9fd979d6-b9b92 1/1 Running 0 10m
coredns-f9fd979d6-t822r 1/1 Running 0 10m
etcd-kubemaster 1/1 Running 0 10m
kube-apiserver-kubemaster 1/1 Running 0 10m
kube-controller-manager-kubemaster 1/1 Running 2 10m
kube-proxy-jpb9p 1/1 Running 0 10m
kube-proxy-lkpv9 1/1 Running 0 6m13s
kube-proxy-sqd9v 1/1 Running 0 6m10s
kube-scheduler-kubemaster 1/1 Running 2 10m
weave-net-8rl49 2/2 Running 0 6m13s
weave-net-fkqdv 2/2 Running 0 6m10s
weave-net-q79pb 2/2 Running 0 7m48s

So, we have a working kubernetes cluster built with kubeadm using vagrant/libvirtd!

As a note, while building the VMs and installing software on them, my laptop hang a couple of times as the 3VMS running at the same time takes nearly all RAM. But this is a good exercise to understand the requirements of kubeadm to build a cluster and as well, it is a lab env you can use while studying if the cloud env are down or you dont have internet. Let’s see If I manage to pass the CKA one day!!!

3VMs running
# top
top - 17:24:10 up 9 days, 18:18, 1 user, load average: 5.22, 5.09, 4.79
Tasks: 390 total, 1 running, 388 sleeping, 0 stopped, 1 zombie
%Cpu(s): 21.7 us, 19.5 sy, 0.0 ni, 56.5 id, 2.0 wa, 0.0 hi, 0.2 si, 0.0 st
MiB Mem : 7867.7 total, 263.0 free, 6798.7 used, 806.0 buff/cache
MiB Swap: 6964.0 total, 991.4 free, 5972.6 used. 409.6 avail Mem
329875 tomas 20 0 9268464 251068 83584 S 55.8 3.1 14:27.84 chrome
187962 tomas 20 0 1302500 105228 46528 S 36.9 1.3 170:58.40 chrome
331127 libvirt+ 20 0 4753296 1.3g 5972 S 35.5 17.5 7:13.00 qemu-system-x86
330979 libvirt+ 20 0 4551524 954212 5560 S 7.3 11.8 4:08.33 qemu-system-x86
5518 root 20 0 1884932 135616 8528 S 5.3 1.7 76:50.45 Xorg
330803 libvirt+ 20 0 4550504 905428 5584 S 5.3 11.2 4:12.68 qemu-system-x86
6070 tomas 9 -11 1180660 6844 4964 S 3.7 0.1 44:04.39 pulseaudio
333253 tomas 20 0 4708156 51400 15084 S 3.3 0.6 1:23.72 chrome
288344 tomas 20 0 2644572 56560 14968 S 1.7 0.7 9:03.78 Web Content
6227 tomas 20 0 139916 8316 4932 S 1.3 0.1 19:59.68 gkrellm

3VMS stopped
root@athens:/home/tomas# top
top - 18:40:09 up 9 days, 19:34, 1 user, load average: 0.56, 1.09, 1.30
Tasks: 379 total, 2 running, 376 sleeping, 0 stopped, 1 zombie
%Cpu(s): 4.5 us, 1.5 sy, 0.0 ni, 94.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
MiB Mem : 7867.7 total, 3860.9 free, 3072.9 used, 933.9 buff/cache
MiB Swap: 6964.0 total, 4877.1 free, 2086.9 used. 4122.1 avail Mem
288344 tomas 20 0 2644572 97532 17100 S 6.2 1.2 11:05.35 Web Content
404910 root 20 0 12352 5016 4040 R 6.2 0.1 0:00.01 top
1 root 20 0 253060 7868 5512 S 0.0 0.1 0:47.82 systemd
2 root 20 0 0 0 0 S 0.0 0.0 0:02.99 kthreadd
3 root 0 -20 0 0 0 I 0.0 0.0 0:00.00 rcu_gp
4 root 0 -20 0 0 0 I 0.0 0.0 0:00.00 rcu_par_gp
6 root 0 -20 0 0 0 I 0.0 0.0 0:00.00 kworker/0:0H
9 root 0 -20 0 0 0 I 0.0 0.0 0:00.00 mm_percpu_wq
10 root 20 0 0 0 0 S 0.0 0.0 0:11.39 ksoftirqd/0
11 root 20 0 0 0 0 I 0.0 0.0 2:13.55 rcu_sched

Kubernetes Troubleshooting I

Restore ETCD

This is a process no well documented in the official docs and I messed up in my CKA exam:

1- check config of etcd process. Maybe you will need some details for the restore process

$ kubectl describe pod -n kube-system etcd-master

2- Stop api-server if not running kubeadm

$ service kube-apiserver stop

3- Check help for all restore options. Keep in mind you will need (very likely) to provide certs for auth.

$ ETCDTL_API=3 etcdctl snapshot restore -h

4- Restore ETCD using a previous backup:

$ ETCDTL_API=3 etcdctl --endpoints snapshot restore FILE \
--cacert xxx --cert xx --key xxx

--data-dir /NEW/DIR \
--initial-cluster-toker TOKEN \ (token is any word) 

--name master \ 
--initial-cluster=master= \ 


5- Add new lines and update volume paths in ETCD config. If it is a static pod, check in /etc/kubernetes/manifests in master node.

--initial-cluster-token TOKEN

++ volumeMounts/volumes to new path /NEW/DIR !!!!

6- Restart services if not running kubeadm

$ systemctl daemon-reload
$ service etcd restart
$ service etcd kube-apiserver start

7- Checks

/// if using kubeadm, docker instance for etcd should restart
$ docker ps -a | grep -i etcd

/// check etcd is running showing members:
$ ETCDCTL_API=3 etcdctl member list --cacert xxx --cert xx --key xxx

Sidecar -logging

Based on this doc. You want to send some logs to stderr so you create a new container that takes those.

Container with a sidecar:

apiVersion: v1
kind: Pod
  name: counter
  - name: count
    image: busybox
   - /bin/sh 
   - -c 
   - > i=0; 
       while true; 
        echo "$i: $(date)" >> /var/log/1.log; 
        echo "$(date) INFO $i" >> /var/log/2.log; i=$((i+1)); sleep 1; 
   - name: varlog 
     mountPath: /var/log
  - name: sidecar-1 
    image: busybox 
    args: [/bin/sh, -c, 'tail -n+1 -f /var/log/1.log'] 
      name: varlog
      mountPath: /var/log
    name: varlog
    emptyDir: {}

Now you can see the logs of “/var/log/1.log” going via “sidecar-1”

$ kubectl logs counter sidecar-1

CPU/Memory of a POD

Based on these links: link1 , link2, link3

If you want to use “kubectl top” you need to install “metrics-server”

$ kubectl top pod --all-namespaces

Keep in mind that “kubectl top” shows metrics for a given pod. That information is based on reports from cAdvisor, which collects real pods resource usage.

And as per link3, “kubectl top” is not the same as running “top” inside the container.

Node NotReady

Based on this link:

$ kubectl get nodes
$ kubectl describe nodes XXX

$ ssh node 
   -> check for kubelet logs 
     cat /var/log/kubelet.log
     $ journalctl -u kubelet // systemctl status kubelet --> if a service


I am studying for the Kubernetes certification CKA. These are some notes:


1.1- Cluster Architecture

Master node: manage, plan, schedule and monitor. These are the main components:

  • etcd: db as k-v
  • scheduler
  • controller-manager: node-controller, replication-controller
  • apiserver: makes communications between all parts
  • docker

Worker node: host apps as containers. Main components:

  • kubelet (captain of the ship)
  • kube-proxy: allow communication between nodes

1.2- ETCD

It is a distributed key-value store (database). TCP 2379. Stores info about nodes, pods, configs, secrets, accounts, roles, bindings etc. Everything related to the cluster.

Basic commands:

client: ./etcdctl set key1 value1
        ./etcdctl get key1 

Install Manual:

1- wget "github binary path to etc"
2- setup config file: important "--advertise-client-urls: IP:2379"
                      a lot of certs needed!!!

Install via kubeadm already includes etcd:

$ kubectl get pods -n kube-system | grep etcd

// get all keys from etcd
$ kubectl exec etcd-master -n kube-system etcdctl get / --prefix -keys-only

etcd can be set up as a cluster, but this is for another section.

1.3- Kube API Server

You can install a binary (like etcd) or use it via kubeadm.

It has many options and it defines certs for all connections!!!

1.4- Kube Controller-Manager

You can install a binary (like etcd) or use kubeadm. It gets all the info via the API server. Watch status of pods, remediate situations. Parts:

  • node-controller
  • replications-controller

1.5- Kube Scheduler

Decides which pod goes to which node. You can install a binary or via kubeadm.

1.6- Kubelet

It is like the “captain” of the “ship” (node). Communicates with the kube-cluster via the api-server.

Important: kubeadm doesnt install kubelet

1.7- Kube-Proxy

In a cluster, each pod can reach any other pod -> you need a pod network!

It runs in each node. Creates rules in each node (iptables) to use “services”

1.8- POD

It is the smallest kube object.

1 pod =~ 1 container + help container

It can be created via a “kubectl run” or via yaml file.

apiVersion: v1
kind: Pod
  name: postgres-pod
    name: postgres-pod
    app: demo-voting-app
    - name: postgres
      image: postgres
        - containerPort: 5432
        - name: POSTGRES_USER
          value: "postgres"
        - name: POSTGRES_PASSWORD
          value: "postgres"


$ kubectl create -f my-pod.yaml
$ kubectl get pods
$ kubectl describe pod postgres

It always contains “apiVersion”, “kind”, “metadata” and “spec”.

1.9 ReplicaSet

Object in charge of monitoring pods, HA, loadbalancing, scaling. It is a replacement of “replication-controller”. Inside the spec.tempate you “cope/paste” the pod definition.

The important part is “selector.matchLabels” where you decide what pods are going to be managed by this replicaset


apiVersion: apps/v1
kind: ReplicaSet
  name: my-rs
    app: myapp
  replicas: 3
  selector: // match pods created before the RS - main difference between RS 
                                                                      and RC
      app: myapp   --> find labels from pods matching this
      name: myapp-pod
        app: myapp
      - name: nginx-controller
        image: nginx


$ kubectl create -f my-rs.yaml
$ kubectl get replicaset
$ kubectl scale --replicas=4 replicaset my-rs
$ kubectl replace -f my-rs.yaml

1.10- Deployments

It is an object that creates a pod + replicaset. It provides the upgrade (rolling updates) feature to the pods.

File is identical as a RS, only changes the “kind”

apiVersion: apps/v1
kind: Deployment
  name: my-deployment
    app: myapp
  replicas: 3
  selector: // match pods created before the RS - main difference between RS 
                                                                   and RC
      app: myapp   --> find labesl from pods matching this
      name: myapp-pod
        app: myapp
      - name: nginx-controller
        image: nginx


$ kubectl create -f my-rs.yaml
$ kubectl get deployments
$ kubectl get replicaset
$ kubectl get pods

1.11- Namespace

It is a way to create different environments in the cluster. ie: production, testing, features, etc. You can control the resource allocations for the “ns”

By default you have 3 namespaces:

  • kube-system: where all control-plane pods are installed
  • default:
  • kube-public:

The “ns” is used in DNS.

---------  --- ---  -----------
svc name   ns  type domain(default)

--------- ---     ---  -----------
pod IP    ns      type  domain(default)

Keep in mind that POD DNS names are just the “IP” in “-” format.

You can add “namespace: dev” into the “metadata” section of yaml files. By default, namespace=default.

$ kubectl get pods --namespace=xx (by default is used "default" namespace)

Create “ns”:

apiVersion: v1
kind: Namespace
name: dev

$ kubectl create -f namespace-dev.yaml

$ kubectl create namespace dev

Change “ns” in your context if you dont want to type it in each kubectl command:

$ kubectl config set-context $(kubectl config current-context) -n dev

See all objects in all ns:

$ kubectl get pods --all-namespaces

$ kubectl get ns --no-headers | wc -l

1.12- Resource Quotas

You can state the resources (cpu, memory, etc) for a pod.


apiVersion: v1
kind: ResourceQuota
 name: compute-quota
 namespace: dev
   pods: "10"
   requests.cpu: "4"
   requests.memory: 5Gi
   limits.cpu: "10"
   limits.memory: 10Gi


$ kubectl create -f compute-quota.yaml

1.13 Services

It is an object. It connects pods to external users or other pods.


  • NodePort: like docker port-mapping
  • ClusterIP: like a virtual IP that is reachable to all pods in the cluster.
  • LoadBalancer: only available in Cloud providers

1.13.1 NodePort

Like a virtual server. SessionAffinity: yes. Random Algorithm for scheduling.

Important parts:

  • targetport: This is the pod port.
  • port: This is the service port (most of the times, it is the same as targetport).
  • nodeport: This is in the node (the port other pods in different nodes are going to hit)


apiVersion: v1
kind: Service
  name: mypapp-service
  type: NodePort
  - targetPort: 80
    port: 80
    nodePort: 30080  (range: 30000-32767)
    app: myapp        ---|
    type: front-end   ---|-> matches pods !!!!

The important bits are the “spec.ports” and “spec.selector” definitions. The “selector” is used to match on labels from pods where we want to apply this service.


// declarative
$ kubectl create -f service-definition.yml
$ kubectl get services

// imperative
$ kubectl expose deployment simple-webapp-deployment --name=webapp-service --target-port=8080 --type=NodePort \
--dry-run=client -o yaml > svc.yaml --> create YAML !!!

Example of creating pod and service imperative way:

$ kubectl run redis --image=redis:alpine --labels=tier=db
$ kubectl expose pod redis --name redis-service --port 7379 --target-port 6379

1.13.2 ClusterIP

It is used for access to several pods (VIP). This is the default service type.


apiVersion: v1
kind: Service
  name: back-end
  type: ClusterIP // (default)
  - targetPort: 80
    port: 80
    app: myapp
    type: back-end


$ kubectl create -f service-definition.yml
$ kubectl get services

1.13.3 Service Bound

Whatever the service you use, you want to be sure it is in use, you can check that seeing if the service is bound to a node. That is configured by “selector” but to confirm that is correct, use the below command. You must have endpoints to proof your service is attached to some pods.

$ kubectl get service XXX | grep -i endpoint

1.13.4 Microservice Architecture Example

Based on this “diagram”:

voting-app     result-app
 (python)       (nodejs)
   |(1)           ^ (4)
   v              |
in-memoryDB       db
 (redis)       (postgresql)
    ^ (2)         ^ (3)
    |             |
    ------- -------
          | |

These are the steps we need to define:

1- deploy containers   -> deploy PODs (deployment)
2- enable connectivity -> create service clusterIP for redis
                          create service clusterIP for postgres
3- external access     -> create service NodePort for voting
                          create service NodePort for result

1.14- Imperative vs Declarative

imperative: how to do things (step by step)

$ kubectl run/create/expose/edit/scale/set …
$ kubectl replace -f x.yaml !!! x.yaml has been updated

declarative: just what to do (no how to do) –> infra as code / ansible, puppet, terraform, etc

$ kublectl apply -f x.yaml <--- it creates/updates

1.15 – kubectl and options

--dry-run: By default as soon as the command is run, the resource will be created. If you simply want to test your command , use the --dry-run=client option. This will not create the resource, instead, tell you weather the resource can be created and if your command is right.

-o yaml: This will output the resource definition in YAML format on screen.

$ kubectl explain pod --recursive ==> all options available

$ kubectl logs [-f] POD_NAME [CONTAINER_NAME]

$ kubectl -n prod exec -it PODNAME cat /log/app.log
$ kubectl -n prod logs PODNAME

1.16- Kubectl Apply

There are three type of files:

  • local file: This is our yaml file
  • live object config: This is the file generated via our local file and it is what you see when using “get”
  • last applied config: This is used to find out when fields are REMOVED from the local file

“kubectl apply” compares the three files above to find our what to add/delete.


2.1- Manual Scheduling

  • what to schedule? find pod without “nodeName” in the spec section, then finds a node for it.
  • only add “nodeName” at creation time
  • After creation, only via API call you can change that

Check you have a scheduler running:

$ kubectl -n kube-system get pods | grep -i scheduler

2.2 Labels and Selectors

  • group and select things together.
  • section “label” in yaml files

how to filter via cli:

$ kubectl get pods --selector key=value --selector k1=v1
$ kubectl get pods --selector key=value,k1=v1
$ kubectl get pods -l key=value -l k1=v1

In Replicasets/Services, the labels need to match!

 replicas: 3
    app:App1 <----
 template:       |
   metadata:     |-- need to match !!!
    labels:      |
     app:App1 <---

2.3 Taints and Tolerations

set restrictions to check what pods can go to nodes. It doesn’t tell the POD where to go!!!

  • you set “taint” in nodes
  • you set “tolerance” in pods


$ kubectl taint nodes NODE_NAME key=value:taint-effect
$ kubectl taint nodes node1 app=blue:NoSchedule <== apply
$ kubectl taint nodes node1 app=blue:NoSchedule- <== remove(-) !!!
$ kubectl taint nodes node1  <== display taints

*tain-effect = what happens to PODS that DO NOT Tolerate this taint? Three types:

- NoSchedule:
- PreferNoSchedule: will try to avoid the pod in the node, but not guarantee
- NoExecute: new pods will not be installed here, and current pods will exit if they dont tolerate the new taint. The node could have already pods before applying the taint…

Apply toleration in pod, in yaml, it is defined under “spec”:

 - key: "app"
   operator: "Equal"
   value: "blue"
   effect: "NoSchedule"

In general, the master node never gets pods (only the static pods for control-plane)

$ kubectl describe node X | grep -i taint

2.4 Node Selector

tell pods where to go (different for taint/toleration)

First, apply on a node a label:

$ kubectl label nodes NODE key=value
$ kubectl label nodes NODE size=Large

Second, apply on pod under “spec” the entry “nodeSelector”:

    size: Large

2.5 Node Affinity

extension of “node selector” with “and” “or” logic ==> mode complex !!!!

apply on pod:#
    requiredDuringSchedulingIgnoredDuringExecution:  or 
      - matchExpressions:
        - key: size
          operator: In    ||   NotIn   ||    Exists
          - Large              Small
          - Medium

DuringScheduling: pod is being created

2.6 Resource Limits

Pod needs by default: cpu(0.5) men(256m) and disk

By default: max cpu = 1 // max mem = 512Mi

Important regarding going over the limit:

if pod uses more cpu than limit -> throttle
                 mem            -> terminate (OOM)


        memory: "1Gi"
        cpu: 1
        memory: "2Gi"
        cpu: 2

2.7 DaemonSets

It is like a replicaset (only kind changes). run 1 pod in each node: ie monitoring, logs viewer, networking (weave-net), kube-proxy!!!

It uses NodeAffinity and default scheduler to schedule pods in nodes.

$ kubectl get daemonset
if you add    a node, the daemonset creates that pod
       delete                       deletes
apiVersion: apps/v1
kind: DaemonSet
  name: monitoring-daemon
      app: monitoring-agent
        app: monitoring-agent
      - name: monitoring-agent
        image: monitoring-agent

2.8 Static PODs

kubelet in a node, can create pods using files in /etc/kubernetes/manifests automatically. But, it can’t do replicasets, deployments, etc

The path for the static pods folder is defined in kubelet config file

kubelet.service <- config file
--config=kubeconfig.yaml \ or

staticPodPath: /etc/kubernetes/manifests

You can check with”docker ps -a” in master for docker images running the static pods.

Static pods is mainly used by master nodes for installing pods related to the kube cluster (control-plane: controller, apiserver, etcd, ..)


  • you can’t delete static pods via kubectl. Only by deleting the yaml file for the folder “/etc/kubernetes/manifests”
  • the pods created via yaml in that folder, will have “-master” added to the name if you are in master node when using “kubectl get pods” or “-nodename” if it is other node.

Comparation Static-Pod vs Daemon-Set

static pod           vs          daemon-set
----------                       -----------
- created by kubelet              - created by kube-api
- deploy control-plane componets  - deploy monitoring, logging
    as static pods                     agents on nodes
- ignored by kube-scheduler       - ignored by kube-scheduler

2.9 Multiple Schedulers

You can write you own scheduler.

How to create it:

--scheduler-name= custom-scheduler

/etc/kubernetes/manifests/kube-scheduler.yam --> copy and modify
--- (a scheduler is a pod!!!)
apiVersion: v1
kind: Pod
  name: my-custom-scheduler
  namespace: kube-system
  - command:
    - kube-scheduler
    - --address=
    - --kubeconfig=/etc/kubernetes/scheduler.conf
    - --leader-elect=false
    - --scheduler-name=my-custom-scheduler
    - --lock-object-name=my-custom-scheduler
    image: xxx
    name: kube-scheduler
    -  containerPort: XXX

Assign new scheduler to pod:

apiVersion: v1
kind: Pod
  name: nginx
  - image: nginx
    name: nginx
  schedulerName: my-custom-scheduler

How to see logs:

$ kubectl get events ---> view scheduler logs
$ kubectl logs my-custom-scheduler -n kube-system


Monitoring cluster components. There is nothing built-in (Oct 2018).

  • pay: datadog, dynatrace
  • Opensource Options: metrics server, prometheus, elastic stack, etc

3.1- metrics server

one per cluster. data kept in memory. kubelet (via cAdvisor) sends data to metric-server.

install: > minukube addons enable metrics-server //or
           other envs: git clone "github path to binary"
                       kubectl create -f deploy/1.8+/

view: > kubectl top node/pod


4.1- Rolling updates / Rollout

rollout -> a new revision. This is the reason you create “deployment” objects.

There are two strategies:

  • recreate: destroy all, then create all -> outage! (scale to 0, then scale to X)
  • rolling update (default): update a container at each time -> no outage (It creates a new replicaset and then starts introducing new pods)

How to apply a new version?

1) Declarative: make change in deployment yaml file
kubectl apply -f x.yaml (recommended)


2) Imperative: 
kubectl create deployment nginx-deploy --image=nginx:1.16
kubectl set image deployment/nginx-deploy nginx=nginx:1.17 --record

How to check status of the rollout

status:   $ kubectl rollout status deployment/NAME
history:  $ kubectl rollout history deployment/NAME
rollback: $ kubectl rollout undo deployment/NAME

4.2- Application commands in Docker and Kube

From a “Dockerfile”:

FROM Ubuntu
ENTRYPOINT ["sleep"] --> cli commands are appended to entrypoint
CMD ["5"] --> if you dont pass any value in "docker run .." it uses by 
              default 5.

With the docker image created above, you can create a container like this:

$ docker run --name ubuntu-sleeper ubuntu-sleeper 10

So now, kubernetes yaml file:

apiVersion: v1
kind: Pod
  name: ubuntu-sleeper-pod
  -  name: ubuntu-sleeper
     image: ubuntu-sleeper
     command: ["sleep","10"] --> This overrides ENTRYPOINT in docker
     args: ["10"]   --> This overrides CMD [x] in docker

4.3- Environment variables

You define them inside the spec.containers.container section:

 - name: x
   image: x
   - containerPort: x
   - name: APP_COLOR
     value: pink

4.4- ConfigMap

Defining env var can be tedious, so config maps is the way to manage them a bit better. You dont have to define in each pod all env vars… just one entry now.

First, create configmap object:

imperative $ kubectl create configmap NAME \
                       --from-literal=KEY=VALUE \
                       --from-literal=KEY2=VALUE2 \
key1: val1
key2: val2

declarative $ kubectl create -f cm.yaml
            $ kubectl get configmaps

cat app-config
apiVersion: v1
kind: ConfigMap
name: app-config

Apply configmap to a container in three ways:

1) Via "envFrom": all vars

  - name: xxx
    envFrom:   // all values
    -  configMapRef:
         name: app-config

2) Via "env", to import only specific vars

 - name: x
   image: x
   - containerPort: x
   - name: APP_COLOR  -- get one var from a configmap, dont import everything
         name: app-config
         key: APP_COLOR

3) Volume:

- name: app-config-volume
    name: app-config

Check “explain” for more info:

$ kubectl explain pods --recursive | grep envFrom -A3

4.5- Secrets

This is encode in base64 so not really secure. It just avoid to have sensitive info in clear text…

A secret is only sent to a node if a pod on that node requires it.
Kubelet stores the secret into a tmpfs so that the secret is not written to disk storage. Once the Pod that depends on the secret is deleted, kubelet will delete its local copy of the secret data as well:

How to create secrets:

imperative $ kubectl create secret generic NAME \
                       --from-literal=KEY=VAL \
cat FILE
DB_Pass: password

declarative $ kubectl create -f secret.yaml

cat secret.yaml
apiVersion: v1
kind: Secret
  name: app-secret
  DB_Pass: HASH <---- $ echo -n 'password' | base64 // ENCODE !!!!
                      $ echo -n 'HASH' | base64 --decode // DECODE !!!!

You can apply secrets in three ways:

1) as "envFrom" to import all params from secret object

  - name: xxx 
    - secretRef:
        name: app-secret

2) Via "env" to declare only one secret param

  - name: x
    image: x
      name: APP_COLOR
          name: app-secret
          key: DB_password

3) Volumes:

  - command: ["sleep", "4800"]
    image: busybox
    name: secret-admin
    - name: secret-volume
      mountPath: "/etc/secret-volume"
      readOnly: true
  - name: secret-volume
      secretName: app-secret --> each key from the secret file is created
                                 as a file in the volume.
                                 The content of the file is the secret.

$ ls -ltr /etc/secret-volume

4.6- Multi-container Pods

Scenarios where your app needs an agent, ie: web server + log agent

apiVersion: v1
kind: Pod
  name: simple-webapp
    name: simple-webapp
 - name: simple-webapp
   image: simple-webapp
   - containerPort: 8080
 - name: log-agent
   image: log-agent

4.7- Init Container

You use an init container when you want to setup something before the other containers are created. Once the initcontainers complete their job, the other containers are created.

An initContainer is configured in a pod like all other containers, except that it is specified inside a initContainers section

You can configure multiple such initContainers as well, like how we did for multi-pod containers. In that case each init container is run one at a time in sequential order.


apiVersion: v1
kind: Pod
  name: myapp-pod
    app: myapp
  - name: init-myservice
    image: busybox
    command: ['sh', '-c', 'git clone <some-repository-that-will-be-used-by-application> ;']
  - name: myapp-container
    image: busybox:1.28
    command: ['sh', '-c', 'echo The app is running! && sleep 3600']


5.1- Drain Node

If you need to upgrade/reboot a node, you need to move the pods to somewhere else to avoid an outage.


$ kubectl drain NODE -> pods are moved to another nods and it doesnt 
                           receive anything new
$ kubectl uncordon NODE -> node can receive pods now

$ kubectl cordon NODE -> it doesnt drain the node, it just make the node to not receive new pods

“kube-controller-manager” check the status of the nodes. By default, kcm takes 5 minutes to mark down:

$ kube-controller-manager --pod-eviction-timeout=5m0s (by default) time masters waits for a node to be backup

5.2- Kubernetes upgrade

You need to check the version you are running:

$ kubectl get nodes --> version: v_major.minor.path

Important: kube only supports only the last two version from current, ie:

new current v1.12 -> support v1.11 and v1.10 ==> v1.9 is not supported!!!

Important: nothing can be higher version than kube-apiserver, ie:

kube-apiserver=x (v1.10)
- controller-mamanger, kube-scheduler can be x or x-1 (v1.10 , v1.9)
- kubetet, kube-proxy can be x, x-1 or x-2 (v1.8, v1.9, v1.10)
- kubectl can be x+1,x,x-1 !!!

Upgrade path: one minor upgrade at each time: v1.9 -> v1.10 -> v1.11 etc

Summary Upgrade:

1- upgrade master node
2- upgrade worker nodes (modes)
- all nodes at the same time
- one node at each time
- add new nodes with the new sw version, move pods to it, delete old node

5.2.1- Upgrade Master

From v1.11 to v1.12

$ kubeadm upgrade plan --> it gives you the info the upgrade

$ apt-get update

$ apt-get install -y kubeadm=1.12.0-00

$ kubeadm upgrade apply v1.12.0

$ kubectl get nodes (it gives you version of kubelet!!!!)

$ apt-get upgrade -y kubelet=1.12.0-00 // you need to do this if you have "master" in "kubectl get nodes"

$ systemctl restart kubelet

$ kubectl get nodes --> you should see "master" with the new version 1.12

5.2.2- Upgrade Worker

From v1.11 to v1.12

master:                     node-1
---------------------       -----------------------
$ kubectl drain node-1
                            apt-get update
                            apt-get install -y kubeadm=1.12.0-00
                            apt-get install -y kubelet=1.12.0-00
                            kubeadm upgrade node \
                                 [config --kubelet-version v1.12.0]
                            systemctl restart kubelet
$ kubectl uncordon node-1
$ apt-mark hold package

5.3- Backup Resources

$ kubectl get all --all-namespaces -o yaml > all-deploy-service.yaml

There are other tools like “velero” from Heptio that can do it. Out of scope for CKA.

5.4- Backup/Restore ETCD – Difficult

“etcd” is important because stores all cluster info.

The difficult part is to get the certificates parameters to get the etcd command working.

– You can get some clues from the static pod definition of etcd:

/etc/kubernetes/manifests/etcd.yaml: Find under exec.command

– or do a ps -ef | grep -i etcd and see the parameters used by other commands

verify command:
ETCDCTL_API=3 etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt \
                      --cert=/etc/kubernetes/pki/etcd/server.crt \
                      --key=/etc/kubernetes/pki/etcd/server.key \
                      --endpoints= member list

create backup:
ETCDCTL_API=3 etcdctl snapshot save SNAPSHOT-BACKUP.db \
                    --endpoints= \
                    --cacert=/etc/etcd/ca.crt \
                    --cert=/etc/etcd/etcd-server.crt \

verify backup:
ETCDCTL_API=3 etcdctl --cacert=/etc/kubernetes/pki/etcd/ca.crt \
                      --cert=/etc/kubernetes/pki/etcd/server.crt \
                      --key=/etc/kubernetes/pki/etcd/server.key \
                      --endpoints= \
                      snapshot status PATH/FILE -w table


etcd backup:
1- documentation: find the basic command for the API version
2- ps -ef | grep etcd --> get path for certificates
3- run command
4- verify backup

5.3.1- Restore ETCD

// 1- Stop api server
$ service kube-apiserver stop

// 2- apply etcd backup
$ ETCDCTL_API=3 etcdctl snapshot restore SNAPSHOT-BACKUP.db \
                  --endpoints= \
                  --cacert=/etc/etcd/ca.crt \
                  --cert=/etc/etcd/etcd-server.crt \
                  --data-dir /var/lib/etcd-from-backup \
                  --initial-cluster master-1=,
                                      master-2=https://x.x.x.y:2380 \
                  --initial-cluster-token NEW_TOKEN \

// 3- Check backup folder
$ ls -ltr /var/lib/etcd-from-backup -> you should see a folder "member"

// 4- Update etcd.service file. The changes will apply immediately as it is a static pod

$ vim /etc/kubernetes/manifests/etcd.yaml
--data-dir=/var/lib/etcd-from-backup (update this line with new path)
--initial-cluster-token=NEW_TOKEN (add this line)volumeMounts:
- mountPath: /var/lib/etcd-from-backup (update this line with new path)
  name: etcd-datavolumes:
- hostPath:
    path: /var/lib/etcd-from-backup (update this line with new path)
    type: DirectoryOrCreate
  name: etcd-data

// 5- Reload services
$ systemctl daemon-reload
$ service etcd restart
$ service kube-apiserver start

Important: In cloud env like aws,gcp you dont have access to ectd…


6.1- Security Primitives

kube-apiserver: who can access: files, certs, ldap, service accounts
                what can they do: RBAC authorization, ABAC autho

6.2- Authentication

Kubectl :

users: admin, devs                   --> kubectl can't create accounts
service accountsL 3rd parties (bots) --> kubectl can create accounts

You can use static file for authentication – NO RECOMMENDED

file x.csv:
   password, user, uid, gid --> --basic-auth-file=x.csv

token token.csv:
   token, user, uid, gid --> --token-auth-file=token.csv

Use of auth files in kube-api config:

  - command: 
    - --basic-auth-file=x.csv 
    // or
    - --token-auth-file=x.csv

Use of auth in API calls:

$ curl -v -k https://master-node-ip:6443/api/v1/pods -u "user1:password1"
$ curl -v -k https://master-node-ip:6443/api/v1/pods \
    --header "Authorization: Bearer TOKEN"

6.3- TLS / Generate Certs

openssl commands to create required files:

gen key:  openssl genrsa -out admin.key 2048
gen cert: openssl rsa -in admin.key -pubout > mybank.pem
gen csr:  openssl req -new -key admin-key -out admin.csr \
                   -subj "/CN=kube-admin/O=system:masters"
             (admin, scheduler, controller-manager, kube-proxy,etc)

Generate cert with SAN:

0) Gen key: 
openssl genrsa -out apiserver.key 2048

1) Create openssl.cnf with SAN info
req_extensions = v3_req
basicConstraints = CA:FALSE
keyUsage = nonRepudiation
subjectAltName = @alt_names
DNS.1 = kubernetes
DNS.2 = kubernetes.default
IP.1 =
IP.2 =

2) Gen CSR:
openssl req -new -key apiserver.key -subj "/CN=kube-apiserve" -out apiserver.csr -config openssl.cnf

3) Sign CSR with CA:
openssl x509 -req -in apiserver.csr -CA ca.crt -CAkey ca.key -out apiserver.crt

Self-Signed Cert: Sign the CSR with own key to generate the cert:

$ openssl x509 -req -in ca.csr -signkey ca.key -out ca.crt

User cers to query API:

$ curl https://kube-apiserver:6443/api/v1/pods --key admin.key --cert admin.crt --cacert ca.crt

Kube-api server config related to certs…:



server cert name => kubelet-nodeX.crt

client cert name => Group: System:Nodes name: system:node:node0x

kubeadm can generate all certs for you:

cat /etc/kubernetes/manifests/kube-apiserver.yaml
  - command:
    - --client-ca-file=
    - --etcd-cafile
    - --etcd-certfile
    - --etcd-keyfile
    - --kubelet-client-certificate
    - --kubelet-client-key
    - --tls-cert-file
    - --tls-private-key-file

How to check CN, SAN and date in cert?

$ openssl x509 -in /etc/kubernetes/pki/apiserver.crt -text -noout

Where you check if there are issues with certs in a core service:

if installed manually: > journalctl -u etcd.service -l
if installed kubeadm: > kubectl logs etcd-master

6.4- Certificates API

Generate certificates is quite cumbersome. So kubernetes has a Certificates API to generate the certs for users, etc

How to create a certificate for a user:

1) gen key for user
openssl genrsa -out new-admin.key 2048

2) gen csr for user
openssl req -new -key new-admin.key -subl "/CN=jane" -out new-admin.csr

3) create "CertificateSigningRequest" kubernetes object:

cat new-admin-csr.yaml
apiVersion: certificates.k8s.io/v1beta1
kind: CertificateSigningRequest
  name: jane
  - system:authenticated
  - digital signature
  - key encipherment
  - server auth
  request: (cat new-admin.csr | base64)

kubectl create -f new-admin-csr.yaml

4) approve new certificate, it can't be done automatically:
kubectl get csr
kubectl certificate approve new-admin

5) show certificate to send to user
kubectl get certificate new-admin -o yaml --> put "certificate:" in (echo ".." | base64 --decode)

The certs used by CA API are in controller-manager config file:


6.5- Kubeconfig

kubectl is always querying the API whenever you run a command and use certs. You dont have to type the certs every time because it is configured in the kubectl config at ~HOME/.kube/config.

The kubeconfig file has three sections: clusters, users and contexts (that join users with contexts). And you can have several of each one.

kubeconfig example:

apiVersion: v1
kind: Config
current-context: dev-user@gcp // example: user@cluster

clusters: ///
  - name:
      certificate-authority: PATH/ca.crt 
      certificate-authority-data: $(cat ca.crt | base64)
      server: https://my-kube-playground:6443

contexts: /// user@cluster
  - name: my-kube-admin@my-kube-playground
    context: my-kube-playground
      user: my-kube-admin
      cluster: my-kube-playground
      namespace: production

users: //
  - name: my-kube-admin
    client-certificate: PATH/admin.crt
    client-key: PATH/admin.key
    client-certificate-data: $(cat admin.crt | base64)
    client-key-data: $(cat admin.key | base64)

You can test other user certs:

$ curl https://kube-apiserver:6443/api/v1/pods --key admin.key \
                                     --cert admin.crt --cacert ca.crt

$ kubectl get pods --server my-kube-playground:6443 \
                   --client-key admin.key \
                   --client-certificate admin.crt \
                   --certificate-authority ca.crt \

Use and view kubeconfig file:

$ kubectl get pods [--kubeconfig PATH/FILE]

$ kubectl config view [--kubeconfig PATH/FILE] <-- show kubectl config file

$ kubectl config use-context prod-user@prod <-- change use-context in file too!

6.6- API groups

This is a basic diagram of the API. Main thing is the difference between “api” (core stuff) and “apis” (named stuff – depends on a namespace):

/metrics  /healthx  /version  /api                /apis          /logs
                             (core)               (named)
                              /v1                   |
                      namespace pods rc      /apps /extensions ... (api groups)
                      pv pvc binding...      /v1                  /v1
                                     /deployments /replicaset  (resources)
                                     -list,get,create,delete,update (verbs)

You can reach the API via curl but using the certs…

$ curl https://localhost:6443 -k --key admin.key --cert admin.crt \
                                 --cacert ca.crt
$ curl https://localhost:6443/apis -k | grep "name"

You can make your life easier using a kubectl proxy that uses the kubectl credentials to access kupeapi

$ kubectl proxy -> launch a proxy in 8001 to avoid use auth each time
                   as it uses the ones from kube config file

$ curl http://localhost:8001 -k


                    kube proxy  != kubeCTL proxy (reach kubeapi)
    (service running on node for 
     pods connectivity)

6.7- Authorization

What you can do. There are several method to arrange authorization:

Node authorizer: (defined in certificate: Group: SYSTEM:NODES CN: system:node:node01)

ABAC (Atribute Base Access Control): difficult to manage. each user has a policy…
{"kind": "Policy", "spec": {"user": "dev-user", "namespace": "", "resource": "pods", "apiGroup": ""}}

RBAC: Role Base Access Control: mode standard usage. create role, assign users to roles

Webhook: use external 3rd party: ie "open policy agent"

AlwaysAllow, AlwaysDeny

You define the method in the kubeapi config file:

--authorization-mode=AlwaysAllow (default)
--authorization-mode=Node,RBAC,Webhook (you use all these mode for each request until allow)

6.8- RBAC

You need to define a role and a binding role (who uses which role) objects. This is “namespaced“.

apiVersion: rbac.authorization.k8s.io/v1
kind: Role
  name: dev
  namespace: xxx
- apiGroups: [""]
  resources: ["pods"]
  verbs: ["list", "get", "create", "update", "delete"]
  resourceNames: ["blue", "orange"] <--- if you want to filter at pod level
                                        too: only access to blue,orange
- apiGroups: [""]
  resources: ["configMap"]
  verbs: ["create"]

$ kubectl create -f dev-role.yaml

apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
  name: dev-binding
  namespace: xxx
- kind: User
  name: dev-user
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: dev
  apiGroup: rbac.authorization.k8s.io

$ kubectl create -f dev-binding.yaml

Info about roles/rolebind:

$ kubectl get roles
          describe role dev
                   rolebinding dev-binding

Important: How to test the access of a user?

$ kubectl auth can-i create deployments [--as dev-user] [-n prod]
                     update pods
                     delete nodes

6.9- Cluster Roles

This is for cluster resources (non-namespae): nodes, pv, csr, namespace, cluster-roles, cluster-roles-binding

You can see the full list for each with:

$ kubectl api-resources --namespaced=true/false

The process is the same, we need to define a cluster role and a cluster role binding:

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
  name: cluster-administrator
- apiGroups: [""]
  resources: ["nodes"]
  verbs: ["list", "get", "create", "delete"]

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
  name: cluster-admin-role-bind
- kind: User
  name: cluster-admin
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: cluster-administrator
  apiGroup: rbac.authorization.k8s.io

Important: You can create a “cluster role” for a user to access pods (ie), using cluster role, that give it access to all pod in all namespaces.

6.10- Images Security

Secure access to images used by pods. An image can be in docker, google repo, etc

image: docker.io/nginx/nginx
           |       |     |
       registry  user  image

from google: gcr.io/kubernetes-e2e-test-images/dnsutils

You can use a private repository:

$ docker login private.io

$ docker run private.io/apps/internal-app

How to define a private registry in kubectl:

kubectl create secret docker-registry regcred \
--docker-server= \
--docker-username= \
--docker-password= \

How to use a specific registry in a pod?

  - name: nginx
    image: private.io/apps/internal-app
      name: regcred

6.11- Security Contexts

Like in docker, you can assign security params (like user, group id, etc) in kube containers. You can set the security params at pod or container level:

at pod level:
  runAsUser: 1000

at container level:
  - name: ubuntu
      runAsUser: 100 (user id)
      capabilities: <=== ONY AT CONTAINER LEVEL!
        add: ["MAC_ADMIN"]

6.12- Network Polices

This is like a firewall, iptables implementation for access control at network level. Regardless the network plugin, all pods in a namespace can reach any other pod (without adding any route into the pod).

Network policies are supported in kube-router, calico, romana and weave-net. It is not supported in flannel (yet)

You have ingress (traffic received in a pod) and egress (traffic generated by a pod) rule. You match the rule to a pod using labels with podSelector:

networkpolicy: apply network rule on pods with label role:db to allow only traffic from pods with label name: api-pod into port 3306

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
  name: db-policy
      role: db
  - Ingress
  - from: 
    - podSelector:
          name: api-pod
    - protocol: TCP
      port: 3306

$ kubectl apply -f xxx

6.13- Commands: kubectx / kubens

I haven’t seen any lab requesting the usage. For the exam is not required but maybe for real envs.

reference: https://github.com/ahmetb/kubectx

With this tool, you don't have to make use of lengthy “kubectl config” commands to switch between contexts. This tool is particularly useful to switch context between clusters in a multi-cluster environment.

sudo git clone https://github.com/ahmetb/kubectx /opt/kubectx
sudo ln -s /opt/kubectx/kubectx /usr/local/bin/kubectx

This tool allows users to switch between namespaces quickly with a simple command.
sudo git clone https://github.com/ahmetb/kubectx /opt/kubectx
sudo ln -s /opt/kubectx/kubens /usr/local/bin/kubens


7.1- Storage in Docker

In docker, /container and /images are under /var/lib/docker.

Docker follows a layered architecture (each line in Dockerfile is a layer):

$ docker build --> Read Only (image layer)
$ docker run -> new layer: it is rw (container layer) - lost once docker finish

So docker follows a “copy-on-write” strategy by default. If you want to be able to access that storage after the docker container is destroyer, you can use volumes:

> docker volume create data_volume 
    --> /var/lib/docker/volumes/data_volume
> docker run -v data_volume:/var/lib/mysql mysql
    --> volume mounting -> dir created in docker folders
> docker run --mount type=bind,source=/data/mysql,target=/var/lib/mysql mysl --> path mounting,dir not created in docker folders

volume driver: local, azure, gce, aws ebs, glusterfs, vmware, etc

storage drivers: enable the layer driver: aufs, zfs, btrfs, device mapper, overlay, overlay2

7.2- Volumes, PersistentVolumes and PV claims.

Volume: Data persistence after container is destroyed

  - image: alpine
    - mountPath: /opt
      name: data-volume ==> /data -> alpine:/opt

  - name: data-volume
      path: /data
      type: Directory

Persistent volumes: cluster pool of volumes that users can request part of it

apiVersion: v1
kind: PersistentVolume
  name: pv-vol1
    - ReadWriteOnce (ReadOnlyMode, ReadWriteMany)
    storage: 1Gi
    path: /tmp/data
  persistentVolumeReclaimPolicy: Retain (default) [Delete, Recycle]

$ kubectl create -f xxx
$ kubectl get persistenvolume [pv]

PV claims: use of a pv. Each pvc is bind to one pv.

apiVersion: v1
kind: PersistentVolumeClaim
  name: myclaim
    - ReadWriteOnce
      storage: 500Mi

$ kubectl create -f xxx
$ kubectl get persistentvolumeclaim [pvc]  
      ==> If status is "bound" you have matched a PV

Use a PVC in a pod:

apiVersion: v1
kind: Pod
  name: mypod
  - name: myfrontend
    image: nginx
    - mountPath: "/var/www/html"
      name: mypd
  - name: mypd
      claimName: myclaim

Important: a PVC will bound to one PV that fits its requirements. Use “get pvc” to check status.

7.3- Storage Class

dynamic provisioning of storage in clouds:

sc-definition -> pvc-definition -> pod-definition 
     ==> we dont need pv-definition! it is created automatically


apiVersion: storage.k8s.io/v1
kind: StorageClass
  name: gcp-storage <===========1
provisioner: kubernetes.io/gce-pd
parameters: (depends on provider!!!!)

apiVersion: v1
kind: PersistentVolumeClaim
  name: myclaim <=========2
  - ReadWriteOnce
  storageClassName: gcp-storage <======1
      storage: 500Mi

use pvc in pod
apiVersion: v1
kind: Pod
  name: mypod
  - name: myfrontend
    image: nginx
    - mountPath: "/var/www/html"
      name: mypd <=======3
  - name: mypd <========3
      claimName: myclaim <===========2


8.1 Linux Networking Basics

$ ip link (show interfaces)

$ ip addr add dev eth0
$ route

$ ip route add via
$ ip route default via

// enabling forwarding
$ echo 1 > /proc/sys/net/ipv4/ip_forward
$ vim /etc/sysctl.conf
  net.ipv4.ip_forward = 1

8.2 Linux DNS basics

$ cat /etc/resolv.conf 
search mycompany.com prod.mycompany.com

$ nslookup x.x.x.x
$ dig

8.3 Linux Namespace

// create ns
ip netns add red
ip netns add blue
ip netns (list ns)
ip netns exec red ip link // ip -n red link
ip netns exec red arp

// create virtual ethernet between ns and assign port to them
ip link add veth-red type veth peer name veth-blue 
  (ip -n red link del veth-red)
ip link set veth-red netns red
ip link set veth-blue netns blue

// assign IPs to each end of the veth
ip -n red addr add dev veth-red
ip -n blue addr add dev veth-blue

// enable links
ip -n red link set veth-red up
ip -n blue link set veth-blue up

// test connectivity
ip netns exec red ping


// create bridge
ip link add v-net-0 type bridge

// enable bridge
ip link set dev v-net-0 up // ( ip -n red link del veth-red)

// create and attach links to bridge from each ns
ip link add veth-red type veth peer name veth-red-br
ip link add veth-blue type veth peer name veth-blue-br

ip link set veth-red netns red
ip link set veth-red-br master v-net-0

ip link set veth-blue netns blue
ip link set veth-blue-br master v-net-0

ip -n red addr add dev veth-red
ip -n blue addr add dev veth-blue

ip -n red link set veth-red up
ip -n blue link set veth-blue up

ip addr add dev v-net-0

ip netns exec blue ip route add via
ip netns exec blue ip route add default via

iptables -t nat -A POSTROUTING -s -j MASQUERADE
iptables -t nat -A PREROUTING --dport 80 --to-destination -j DNAT

8.4 Docker Networking

Three types:

- none: no connectivity
- host: share host network
- bridge: internal network is created and host is attached
   (docker network ls --> bridge -| are the same thing
    ip link --> docker0          -|

iptables -t nat -A DOCKER -j DNAT --dport 8080 --to-destination

8.5 Container Network Interface

Container runtime must create network namespace:
- identify network the container must attach to
- container runtime to invoke network plugin (bridge) when container is added/deleted
- json format of network config

 must support command line arguments add/del/chec
 must support parametes container id, network ns
 manage IP
 resutls in specific format

**docker is not a CNI**

kubernetes uses docker. it is created in the "host" network and then uses "bridge"

8.6 Cluster Networking

Most common ports:

etcd: 2379 (2380 as client)
kube-api: 6443
kubelet: 10250
kube-scheduler: 10251
kube-controller: 10252
services: 30000-32767

Configure weave-network:

$ kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')"

$ kubectl get pod -n kube-system | grep -i weave (one per node)

cluster-networking doc: Doesnt give you steps to configure any CNI….

8.7 Pod Networking

  • every pod should have an ip.
  • every pod shoud be able to community with every other pod in the same node and other nodes (without nat)

Networking config in kubelet:

./net-script.sh add <container> <namespace>

8.8 CNI Weave-net

installs an agent in each node. deploy as pods in nodes

$ kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')" 

$ kubectl get pods -n kube-system | grep weave-net

ipam weave:

where pods and bridges get the IPs?
plugin: host-local -> provide free ips from node

8.9 Service Networking

service” is cluster-wide object. The service has an IP. Kubeproxy in each node, creates iptables rules.

ClusterIP: IP reachable by all pods in the cluster

$ ps -ef | kube-api-server --service-cluster-ip-range=x.x.x.x/y
!! pod network shouldnt overlap with service-cluster
$ iptables -L -t -nat | grep xxx
$ cat /var/log/kube-proxy.log

NodePort: same port in all nodes, sent to the pod

IPs for pod: check logs of pod weave:

$ kubectl -n kube-system logs weave-POD weave 
    --> the pod has two container so you need to specify one of them

IPs for services –> check kube-api-server config

8.10 CoreDNS

For pods and services in the cluster (nodes are managed externally)

kube dns: hostname    namespace  type  root           ip address
          web-service apps       svc   cluster.local  x.x.x.x (service)
          10-244-2-5  default    pod   cluster.local  x.x.x.y (pod)

fqdn: web-service.apps.svc.cluster.local

dns implementation in kubernetes use coredns (two pods for ha)

cat /etc/coredns/Corefile
.53: {
  errors // plugins
  kubernetes cluster.local in-addr.arpa ip6.arpa {
     pods insecure // create record for pod as 10-2-3-1 instead of
     fallthrough in-addr.arpa ip6.arpa
  prometheus: 9153
  proxy: . /etc/resolv.conf // for external queries (google.com) from a pod
  cache: 30

$ kubectl get configmap -n kube-system

pods dns config:

cat /etc/resolv.conf => nameserver IP 
    <- it is the IP from $ kubectl get service -n kubesystem | grep dns
                         this come from the kubelet config:

$ host ONLY_FQDN

8.11 Ingress

Using a service “LoadBalance” is only possible in Cloud env like GCP, AWS, etc

When you create a service loadbalancer, the cloud provider is going to create a proxy/loadbalancer to access that service. so you can create a hierarchy of loadbalancers in the cloud provider… –> too complex ==> sol: Ingress

ingress = controller + resources. Not deployed by default

supported controller: GCP HTTPS Load Balancer (GCE) and NGINX (used in kubernetes)

8.11.1 Controller

1) nginx --> deployment file:
apiVersion: apps/v1
kind: Deployment
  name: nginx-ingress-controller
  replicas: 1
      name: nginx-ingress
        name: nginx-ingress
      - name: nginx-ingress-controller
        image: quay.io/kubernetes-ingress-controller/nginx-ingress-controller:0.21.0
      - /nginx-ingress-controller
      - --configmap=$(POD_NAMESPACE)/nginx-configuration
      - name: POD_NAME
            fieldPath: metadata.name
      - name: POD_SPACE
            fieldPath: metadata.namespace
      - name: http
        containerPort: 80
      - name: https:
        containerPorts: 443

2) nginx configmap used in deployment
apiVersion: v1
kind: ConfigMap
  name: nginx-configuration

3) service
apiVersion: v1
kind: Service
  name: nginx-ingress
  type: NodePort
  - port: 80
    targetPort: 80
    protocol: TCP
    name: http
  - port: 443
    targetPort: 443
    protocol: TCP
    name: https
    name: nginx-ingress

4) service account (auth): roles, clusterroles, rolebinding, etc
apiVersion: v1
kind: ServiceAccount
  name: nginx-ingress-serviceaccount

8.11.2 Options to deploy ingress rules

option1) 1rule/1backend: In this case the selector from the service, gives us the pod

apiVersion: extensions/v1beta1
kind: Ingress
  name: ingress-wear
    serviceName: wear-service
    servicePort: 80

option 2) split traffic via URL: 1 Rule / 2 paths

          /wear              /watch
           |                     |
          svc                   svc
          wear                  vid
          ====                  ====
           |                      |
        wear-pod               vid-pod

apiVersion: extensions/v1beta1
kind: Ingress
  name: ingress-wear-watch
  - http: 
      - path: /wear
          serviceName: wear-service
          servicePort: 80
      - path: /watch
          serviceName: watch-service
          servicePort: 80

$ kubectl describe ingress NAME
    ==> watchout the default backend !!!! 
        if nothing matches, it goes there!!!
        you need to define a default backend

option 3) split by hostname: 2 Rules / 1 path each

wear.my-online-store.com           watch.my-online-store.com
                |                    |
               svc                  svc
               wear                 vid
               ====                 ====
                |                    |
            wear-pod               vid-pod

apiVersion: extensions/v1beta1
kind: Ingress
  name: ingress-wear-watch
  - host: wear.my-online-store.com 
      - backend:
          serviceName: wear-service
          servicePort: 80
  - host: watch.my-online-store.com
      - backend:
          serviceName: watch-service
          servicePort: 80

ingress examples: https://kubernetes.github.io/ingress-nginx/examples/

8.12 Rewrite

I havent seen any question about this in the mock labs but just in case: Rewrite url nginx:

For example: replace(path, rewrite-target)
using: http://<ingress-service>:<ingress-port>/wear 
   --> http://<wear-service>:<port>/

In our case: replace("/wear","/")

apiVersion: extensions/v1beta1
kind: Ingress
  name: test-ingress
  namespace: critical-space
    nginx.ingress.kubernetes.io/rewrite-target: /
  - http: 
      - path: /wear
          serviceName: wear-service
          servicePort: 8282

with regex
replace("/something(/|$)(.*)", "/$2")

apiVersion: extensions/v1beta1
kind: Ingress
    nginx.ingress.kubernetes.io/rewrite-target: /$2
  name: rewrite
  namespace: default
  - host: rewrite.bar.com 
      - backend:
          serviceName: http-svc
          servicePort: 80
        path: /something(/|$)(.*)

9- Troubleshooting

9.1 App failure

- make an application diagram
- test the services: curl, kubectl describe service (compare with yaml)
- status pod (restarts), describe pod, pod logs (-f)

9.2 Control plane failure

- get nodes, get pods -n kube-system

- master: service kube-apiserver/kube-controller-manager/
                                            kube-scheduler  status
          kubeadm: kubectl logs kube-apiserver-master -n kube-system
          service: sudo journalctl -u kube-apiserver

- worker: service kubelet/kube-proxy status

- Do exist static pods configured in kubelet config?
   1 check /etc/systemd/system/kubelet.service.d/10-kubeadm.confg for config file
   2 check static pod path in kubelet config

9.3 Worker node failure

- get nodes, describe nodes x (check status column)
- top, dh, service kubelet status, kubelet certificates, kubelet service running?
- kubectl cluster-info


10.1 Basics

$ = root dictionary
results are always in [] // list

$.car.price -> [1000]
  "car": {
    "color": "blue",
    "price": "1000"
  "bus": {
    "color": "red",
    "price": "1200"

$[0] -> ["car"]

$[?(@>40)] == get all numbers greater than 40 in the array -> [45, 60]

$.car.wheels[?(@.location == "xxx")].model

// find prize winner named Malala
$.prizes[?(@)].laureates[?(@.firstname == "Malala")]


find the first names of all winners of year 2014
$.prizes[?(@.year == 2014)].laureates[*].firstname

$[0:3] (start:end) -> 0,1,2 (first 3 elements)
$[0:8:2] (start:end:step) -> 0,0+2=2,2+2=4,4+2=6 -> 
                                elements in position 0,2,4,6
$[-1:0] = last element
$[-1:] = last element
$[-3:] = last 3 elements

10.2 Jsonpath in Kubernetes

$ kubectl get pods -o json

$ kubectl get nodes -o=jsonpath='{.items[*].metada.name}{"\n"}
master node01
4      4

$ kubectl get nodes -o=jsonpath='{range .items[*]}\
master 4
node01 4

$ kubectl get nodes -o=custom-columns=NODE:.metadata.name,
                                      CPU:.status.capacity.cpu …
master 4
node01 4

$ kubectl get nodes --sort-by= .metadata.name

$ kubectl config view --kubeconfig=/root/my-kube-config 
            -o=jsonpath='{.users[*].name}' > /opt/outputs/users.txt

$ kubectl config view --kubeconfig=my-kube-config 
       -o jsonpath="{.contexts[?(@.context.user=='aws-user')].name}" >


11- Install, Config and Validate Kube Cluster

All based on this.

11.1- Basics

education: minikube

on-prem: kubeadm

laptop: minikube: deploys VMs (that are ready) - single node cluster
        kubeadm: require VMS to be ready - single/multi node cluster

turnkey solution: you provision, configure and maintein VMs. 
                  Use scripts to deploy cluster (KOPS in AWS)
                 ie: openshift (redhat), Vagrant, VMware PKS, Cloud Foundry

hosted solutions: (kubernetes as a service) provider provision and maintains VMs, install kubernetes: ie GKE in GCP

11.2 HA for Master

api-server --> need LB (active-active)

$ kube-controller-manager --leader-elect true [options]
  --leader-elect-lease-duration 15s
  --leader-elect-renew-deadline 10s
  --leader-elect-retry-period 2s

etcd: inside the masters (2 nodes total) or in separated nodes (4 nodes total)

11.3 HA for ETCD

leader etcd, writes and send the info to the others
leader election - RAFT:
   quorum = n/2 + 1 -> minimun num of nodes to accept a transactio
   recommend: 3 etcd nodes minimun => ODD NUMBER

$ export ETCDCTL_API=3
$ etcdctl put key value
$ etcdctl get key
$ etcdctl get / --prefix --keys-only

11.4 Lab Deployment

LAB setup (5nodes)
  1 LB
  2 master nodes (with etcd)
  2 nodes

> download kubernetes latest release from github
> uncompress
> cd kubernetes
> cluster/get-kube-binaries.sh --> downloads the latest binaries for your system.
> cd server; tar -zxvf server-linux-xxx
> ls kubernetes/server/bin

1- deploy etcd cluster
2- deploy control plane components (api-server, controller-manager, scheduler)
3- configure haproxy (for apiserver)

 |                       |
 M1:                     M2:
 api                     api
 etcd                    etcd
 control-manager         control-manager
 scheduler               scheduler

 W1:                      W2:
 gen certs                TLS Bootstrap:
 config kubelet             - w2 creates and configure certs itself
 renew certs                - config kubelet
 config kube-proxy          - w2 to renew certs by itself
                            - config kube-proxy

TLS bootstrap:
1- in Master
 - create bootstrap token and associate it to group "system:bootstrappers"
 - assign role "system:node-bootstrapper" to group "system:bootstrappers"
 - assing role "system:certificates.k8s.io:certificatesigningrequests:nodeclient" to group "system:bootstrappers"
 - assing role "system:certificates.k8s.io:certificatesigningrequests:selfnodeclient" to group "system:node"

2- kubelet.service
       // This is for getting the certs to join the cluster!!
   --rotate-certificates=true // this if for the client certs used to join the cluster (CSR automatic approval)
   --rotate-server-certificates=true // these are the certs we created in the master and copied to the worker manually
the server cert requires CSR manual approval !!!

> kubectl get csr
> kubectl certificate approve csr-XXX

apiVersion: 1
- cluster:
    certificate-authority: /var/lib/kubernetes/ca.crt
    server: //(api-server lb IP)
  name: bootstrap
- context:
    cluster: bootstrap
    user: kubelet-bootstrap
  name: bootstrap
current-context: bootstrap
kind: Config
preferences: {}
- name: kubelet-bootstrap
    token: XXXXXXXXXX

11.5 Testing

11.5.1 manual test

$ kubectl get nodes
              pods -n kube-system (coredns, etcd, kube-paiserver, controller-mamanger, proxy, scheduler, weave)

$ service kube-apiserver status

$ kubectl run nginx
          get pods
          scale --replicas=3 deploy/nginx
          get pods

$ kubectl expose deployment nginx --port=80 --type=NodePort
          get service
$ curl http://worker-1:31850

11.5.2 kubetest

end to end test: 1000 tests (12h) // conformance: 160 tests (1.5h)

1- prepare: creates a namespace for this test
2- creates test pod in this namespace, waits for the pods to come up
3- test: executes curl on one popd to reach the ip of another pod over http
4- record result

$ go get -u k8s.io/test-infra/kubetest
$ kubetest --extract=v1.11.3 (your kubernetes version)
$ cd kubernetes
$ export KUBE_MASTER_IP=""
$ export KUBE_MASTER=kube-master
$ kubetest --test --provider=skeleton > test-out.txt // takes 12 hours
$ kubetest --test --provider=skeleton --test_args="--ginkgo.focus=[Conformance]" > testout.txt // takes 1.5 hours

$ kubeadm join --token vab2bs.twzblu86r60qommq \
--discovery-token-ca-cert-hash sha256:3c9b88fa034a6f894a21e49ea2e2d52435dd71fa5713f23a7c2aaa83284b6700

12- Official cheatsheet