I have just started experimenting with K8S a few days back, try to learn K8S with specific emphasis on networking, service mesh etc.
I am running 2 worker nodes and 1 master on VMs with Centos 7 and K8S, installed with kubeadm.
Default CNI of Flannel. Install was OK and everything except the networking was working. I could deploy containers etc, so a lot of control plane was working.
However, networking not working correctly, even container to container in same worker node. I checked all the usual suspects, the veths, IPs, MACs, briges on a single worker and everything seemed to check out... e.g. MACs where on the correct bridges i.e. cni0, IP address assignments etc. Even when pinging from busybox to busybox, I would see the ARP caches being populated but pings not working still.... disabled all FWs, IP forwarding enabled etc. Not an expert of IPtables but looked OK..... also when logged into the worker node shell I could ping the busybox containers, but they could not ping each other....
One question I have at this point, why is the docker0 bridge still present even when flannel is installed can I delete it or are there some dependencies associated with it ? I did not notice the veths for the containers were showing connected to docker0 bridge but docker bride0 was down... however I followed this website and it show a different way of validating and show veths connected to cni0, which is very confusing and frustrating.....
I gave up Flannel as I was just using Flannel to experiment and decided to try out Calico....
I followed install procedures from Calico site... was not entirely clear on the tidy up procedures for Flannel, not sure where this is documented?... this is where it went from bad to worse...
I started getting crash loops on calico containers and coredns stuck creating, reporting liveliness issues on calico ....and this is where I am stuck......... and would like some help.......
I am have read and tried many things on web and may have fixed some issues as there may be many in play, but would really appreciate any help....
=== install info and some output...
enter image description here
enter image description here
enter image description here
enter image description here
enter image description here
Some questions...
The Container creating for the coredns..... is this dependent on successful install of Calico... are the issues related.... or should coredns install work independent of the CNI install ?
The Container creating for the coredns..... is this dependent on successful install of Calico... are the issues related.... or should coredns install work independent of the CNI install ?
Yes, it is. You need to install a CNI to have coredns working.
When you setup your cluster with kubeadm there's is a flag called --pod-network-cidr, depending on which CNI you intend to use, you need to specify the range with this flag.
For example, by default, Calico use the range 192.168.0.0/16 and Flannel use the range 10.244.0.0/16.
I have a guide how to setup a cluster using kubeadm, maybe it help you.
Please note, if you want to replace the CNI without delete the entire cluster, extras steps need to be taken in order to "cleanup" the firewall rules from the older CNI.
See here how to replace flannel to calico, for example.
And here how to migrate from flannel to calico.
Related
I am trying to install a Kubernetes cluster with one master node and two worker nodes.
I acquired 3 VMs for this purpose running on Ubuntu 21.10. In the master node, I installed kubeadm:1.21.4, kubectl:1.21.4, kubelet:1.21.4 and docker-ce:20.4.
I followed this guide to install the cluster. The only difference was in my init command where I did not mention the --control-plane-endpoint. I used calico CNI v3.19.1 and docker for CRI Runtime.
After I installed the cluster, I deployed minio pod and exposed it as a NodePort.
The pod got deployed in the worker node (10.72.12.52) and my master node IP is 10.72.12.51).
For the first two hours, I am able to access the login page via all three IPs (10.72.12.51:30981, 10.72.12.52:30981, 10.72.13.53:30981). However, after two hours, I lost access to the service via 10.72.12.51:30981 and 10.72.13.53:30981. Now I am only able to access the service from the node on which it is running (10.72.12.52).
I have disabled the firewall and added calico.conf file inside /etc/NetworkManager/conf.d with the following content:
[keyfile]
unmanaged-devices=interface-name:cali*;interface-name:tunl*;interface-name:vxlan.calico
What am I missing in the setup that might cause this issue?
This is a community wiki answer posted for better visibility. Feel free to expand it.
As mentioned by #AbhinavSharma the problem was solved by switching from Calico to Flannel CNI.
More information regarding Flannel itself can be found here.
Since yesterday, I am struggling with this strange issue: node "kmaster" not found.
I tried with multiple combinations of installing kubernetes on jetstream instance.
using calico in ubuntu
using flannel in centos
and few other ways
I looked it online and found many people have the same issue:
https://github.com/kubernetes/kubernetes/issues/61277
If someone ran into a similar issue, then please let me know what steps are needed to be taken to resolve it.
Thanks.
I would recommend to bootstrap Kubernetes cluster from scratch and share with you some helpful links with steps how to proceed:
Kubernetes cluster install on Ubuntu with Calico CNI
Kubernetes cluster install on Centos with Flannel CNI
Keep in mind to fulfill system requirements before you start with kubeadm installation procedure.
You can also take a look at the general kubeadm installation or runtime troubleshooting guide.
I have found my solution for this. I was having issue running kubernetes cluster because the kubernetes components are distributed on multiple servers. Once I created the master node and slave(worker) node on the same machine, the issue got resolved.
The steps that I took to resolve the issue:
1. on slave/worker machine, run this command: kubeadm reset
2. on master node, generate the token by running this command: kubeadm generate token.
3. use the token generated in master machine on the slave node, so that the node machine can join the kubernetes cluster.
Cheers!!
I have been setting up Kubernets with kubeadm and I have used Flannel to setup the pod network. The setup basically worked but I have been running into all kinds of problems (and bugs) and now I am trying to gain a better understanding of the different steps involved in network setup process (e.g. CNI and flannel).
From an end-user/admin perspective I simply pass --pod-network-cidr with some network argument to kubeadm and then later I apply a pod configuration for flannel using kubectl. Kubernetes will then start a flannel pod on each of my nodes. Assuming everything worked, flannel should then use the container network interfaces (CNI) of Kubernetes to setup the pod network.
As a result of this process I should get a pod network which includes the following:
A cni0 bridge.
A flannel.x interface.
iptables entries to route between the host and the pod network.
The following files and binaries seem to be involved in the setup:
kubectl reads a CNI configuration such as /etc/cni/net.d/10-flannel.conflist and invokes the CNI plugin described in the config file.
Somehow a folder /var/lib/cni is being created which seems to contain configuration files for the network setup.
A CNI plugin such as /opt/cni/bin/flannel is run, I don't yet understand what it does.
What am I missing on this list and how does (2.) fit into these steps. How does /var/lib/cni get created and which program is responsible for this?
As I see from code of CNI:
var (
CacheDir = "/var/lib/cni"
)
this folder used as cache dir for CNI and looks like created by CNI plugin.
Here you can find detailed documentation about CNI.
What is CNI?
CNI (Container Network Interface), a Cloud Native Computing Foundation project, consists of a specification and libraries for writing plugins to configure network interfaces in Linux containers, along with a number of supported plugins. CNI concerns itself only with network connectivity of containers and removing allocated resources when the container is deleted. Because of this focus, CNI has a wide range of support and the specification is simple to implement.
All projects like Calico, Flannel use CNI as a base. That's why they called CNI-plugins
Here you can find documentation about how kubernetes interact with CNI.
after using k8s on GKE for a couple of months I've decided to install my own cluster. Now, I've two ubuntu VMs and one of them is the kube-master and the second one is a node. Cluster run as I expect and I can see nodes (kube-master is a node also) when run kubectl get nodes. I've launched one pod each VMs bat I'm experiencing an issue both pod have the same IP. Can anybody help me to resolve this issue? I'm using flannel as network plugin atm.
Thanks in advance
Update
I've found the solution, thanks kubernetes group on slack. I didn't install the [cni][1]plugin so the kubelet didn't know the subnetwork status. I installed the plugin using this guide, made a configuration file following that. Restarted the kubelete service finally I saw the cluster was work as I expected.
While getting familiar with kubernetes I do see tons of tools that should helps me to install kubernetes anywhere, but I don't understand exactly what it does inside, and as a result don't understand how to trouble shoot issues.
Can someone provide me a link with tutorial how to install kubernetes without any tools.
There are two good guides on setting up Kubernetes manually:
Kelsey Hightower's Kubernetes the hard way
Kubernetes guide on getting started from scratch
Kelsey's guide assumes you are using GCP or AWS as the infrstructure, while the Kubernetes guide is a bit more agnostic.
I wouldn't recommend running either of these in production unless you really know what you're doing. However, they are great for learning what is going on under the hood. Even if you just read the guides and don't use them to setup any infrastructure you should gain a better understanding of the pieces that make up a Kubernetes cluster. You can then use one of the helpful setup tools to create your cluster, but now you will understand what it is actually doing and can debug when things go wrong.
For simplicity, you can view k8s as three components
etcd
k8s master, which includes kube-apiserver, controller, scheduler
node, which contains kubelet
You can install etcd and k8s master together in one machine. The procedures are
Install etcd. Download etcd package and run it, which is quite
simple. Remember the port of etcd service, e.g. 2379,4001, or any you
set.
Git clone the kubernetes project from github. Find the executable binary file, e.g. for k8s version 1.3, you can find kube-apiserver, kube-controller-manager and kube-scheduler in src/k8s.io/kubernetes/_output/local/bin/linux/amd64 folder
Then run kube-apiserver, specify the etcd ip and port (e.g. --etcd_servers=http://127.0.0.1:4001)
Run scheduler and controller, specifying the apiserver ip and port(e.g. --master=127.0.0.1:8080). There is no oreder between scheduler and controller
Master is running so far. Make sure these processes run without errors. If etcd exits, apiserver would exit. If apiserver exits, scheduler and controller would exit.
On another machine(virtual preferred, network connected), run kubelet. Kubelet could also be found in previous folder(src/k8s.io/kubernetes/_output/local/bin/linux/amd64), specify apiserver ip and port(e.g. --api-servers=http://10.10.10.19:8080). You may install docker or something else on node, which to prove that you could create a container.