KIND (Kubernetes in Docker) deployment of OVN kubernetes is a fast and easy means to quickly install and test kubernetes with OVN kubernetes CNI. The value proposition is really for developers who want to reproduce an issue or test a fix in an environment that can be brought up locally and within a few minutes.
- 20 GB of free space in root file system
- Docker run time or podman
- KIND
-
Installation instructions can be found at https://github.com/kubernetes-sigs/kind#installation-and-usage.
-
NOTE: The OVN-Kubernetes ovn-kubernetes/contrib/kind.sh and ovn-kubernetes/contrib/kind.yaml files provision port 11337. If firewalld is enabled, this port will need to be unblocked:
sudo firewall-cmd --permanent --add-port=11337/tcp; sudo firewall-cmd --reload
-
- kubectl
- Python and pip
- jq
NOTE : In certain operating systems such as CentOS 8.x, pip2 and pip3 binaries are installed instead of pip. In such situations create a softlink for "pip" that points to "pip2".
For OVN kubernetes KIND deployment, use the kind.sh script.
First Download and build the OVN-Kubernetes repo:
$ go env -w GO111MODULE=auto
$ go get github.com/ovn-org/ovn-kubernetes; cd $(go env GOPATH)/src/github.com/ovn-org/ovn-kubernetes
The kind.sh script builds OVN-Kubernetes into a container image. To verify
local changes before building in KIND, run the following:
$ pushd go-controller
$ make
$ popd
$ pushd dist/images
$ make fedora
$ popd
Launch the KIND Deployment.
$ pushd contrib
$ export KUBECONFIG=${HOME}/ovn.conf
$ ./kind.sh
$ popd
This will launch a KIND deployment. By default the cluster is named ovn.
$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
ovn-control-plane Ready master 5h13m v1.16.4
ovn-worker Ready <none> 5h12m v1.16.4
ovn-worker2 Ready <none> 5h12m v1.16.4
$ kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system coredns-5644d7b6d9-kw2xc 1/1 Running 0 5h13m
kube-system coredns-5644d7b6d9-sd9wh 1/1 Running 0 5h13m
kube-system etcd-ovn-control-plane 1/1 Running 0 5h11m
kube-system kube-apiserver-ovn-control-plane 1/1 Running 0 5h12m
kube-system kube-controller-manager-ovn-control-plane 1/1 Running 0 5h12m
kube-system kube-scheduler-ovn-control-plane 1/1 Running 0 5h11m
local-path-storage local-path-provisioner-7745554f7f-9r8dz 1/1 Running 0 5h13m
ovn-kubernetes ovnkube-db-5588bd699c-kb8h7 2/2 Running 0 5h11m
ovn-kubernetes ovnkube-master-6f44d456df-bv2x8 3/3 Running 0 5h11m
ovn-kubernetes ovnkube-node-2t6m2 3/3 Running 0 5h11m
ovn-kubernetes ovnkube-node-hhsmk 3/3 Running 0 5h11m
ovn-kubernetes ovnkube-node-xvqh4 3/3 Running 0 5h11m
The kind.sh script defaults the cluster to HA disabled. There are numerous
configuration options when deploying. Use ./kind.sh -h to see the latest options.
[root@ovnkubernetes contrib]# ./kind.sh --help
usage: kind.sh [[[-cf |--config-file <file>] [-kt|keep-taint] [-ha|--ha-enabled]
[-ho |--hybrid-enabled] [-ii|--install-ingress] [-n4|--no-ipv4]
[-i6 |--ipv6] [-wk|--num-workers <num>] [-ds|--disable-snat-multiple-gws]
[-dp |--disable-pkt-mtu-check]
[-nf |--netflow-targets <targets>] [sf|--sflow-targets <targets>]
[-if |--ipfix-targets <targets>] [-ifs|--ipfix-sampling <num>]
[-ifm|--ipfix-cache-max-flows <num>] [-ifa|--ipfix-cache-active-timeout <num>]
[-sw |--allow-system-writes] [-gm|--gateway-mode <mode>]
[-nl |--node-loglevel <num>] [-ml|--master-loglevel <num>]
[-dbl|--dbchecker-loglevel <num>] [-ndl|--ovn-loglevel-northd <loglevel>]
[-nbl|--ovn-loglevel-nb <loglevel>] [-sbl|--ovn-loglevel-sb <loglevel>]
[-cl |--ovn-loglevel-controller <loglevel>]
[-ep |--experimental-provider <name>] |
[-eb |--egress-gw-separate-bridge]
[-h]]
-cf | --config-file Name of the KIND J2 configuration file.
DEFAULT: ./kind.yaml.j2
-kt | --keep-taint Do not remove taint components.
DEFAULT: Remove taint components.
-ha | --ha-enabled Enable high availability. DEFAULT: HA Disabled.
-ho | --hybrid-enabled Enable hybrid overlay. DEFAULT: Disabled.
-ds | --disable-snat-multiple-gws Disable SNAT for multiple gws. DEFAULT: Disabled.
-dp | --disable-pkt-mtu-check Disable checking packet size greater than MTU. Default: Disabled
-nf | --netflow-targets Comma delimited list of ip:port or :port (using node IP) netflow collectors. DEFAULT: Disabled.
-sf | --sflow-targets Comma delimited list of ip:port or :port (using node IP) sflow collectors. DEFAULT: Disabled.
-if | --ipfix-targets Comma delimited list of ip:port or :port (using node IP) ipfix collectors. DEFAULT: Disabled.
-ifs | --ipfix-sampling Fraction of packets that are sampled and sent to each target collector: 1 packet out of every <num>. DEFAULT: 400 (1 out of 400 packets).
-ifm | --ipfix-cache-max-flows Maximum number of IPFIX flow records that can be cached at a time. If 0, caching is disabled. DEFAULT: Disabled.
-ifa | --ipfix-cache-active-timeout Maximum period in seconds for which an IPFIX flow record is cached and aggregated before being sent. If 0, caching is disabled. DEFAULT: 60.
-el | --ovn-empty-lb-events Enable empty-lb-events generation for LB without backends. DEFAULT: Disabled
-ii | --install-ingress Flag to install Ingress Components.
DEFAULT: Don't install ingress components.
-n4 | --no-ipv4 Disable IPv4. DEFAULT: IPv4 Enabled.
-i6 | --ipv6 Enable IPv6. DEFAULT: IPv6 Disabled.
-wk | --num-workers Number of worker nodes. DEFAULT: HA - 2 worker
nodes and no HA - 0 worker nodes.
-sw | --allow-system-writes Allow script to update system. Intended to allow
github CI to be updated with IPv6 settings.
DEFAULT: Don't allow.
-gm | --gateway-mode Enable 'shared' or 'local' gateway mode.
DEFAULT: shared.
-ov | --ovn-image Use the specified docker image instead of building locally. DEFAULT: local build.
-ml | --master-loglevel Log level for ovnkube (master), DEFAULT: 5.
-nl | --node-loglevel Log level for ovnkube (node), DEFAULT: 5
-dbl | --dbchecker-loglevel Log level for ovn-dbchecker (ovnkube-db), DEFAULT: 5.
-ndl | --ovn-loglevel-northd Log config for ovn northd, DEFAULT: '-vconsole:info -vfile:info'.
-nbl | --ovn-loglevel-nb Log config for northbound DB DEFAULT: '-vconsole:info -vfile:info'.
-sbl | --ovn-loglevel-sb Log config for southboudn DB DEFAULT: '-vconsole:info -vfile:info'.
-cl | --ovn-loglevel-controller Log config for ovn-controller DEFAULT: '-vconsole:info'.
-ep | --experimental-provider Use an experimental OCI provider such as podman, instead of docker. DEFAULT: Disabled.
-eb | --egress-gw-separate-bridge The external gateway traffic uses a separate bridge.
--delete Delete current cluster
As seen above, if you do not specify any options the script will assume the default values.
To verify local changes, the steps are mostly the same as with docker, except the fedora make target:
$ OCI_BIN=podman make fedora
To deploy KIND however, you need to start it as root and then copy root's kube config to use it as non-root:
$ pushd contrib
$ sudo ./kind.sh -ep podman
$ sudo cp /root/ovn.conf ~/.kube/kind-config
$ sudo chown $(id -u):$(id -g) ~/.kube/kind-config
$ export KUBECONFIG=~/.kube/kind-config
$ popd
Notes / troubleshooting:
- Issue with /dev/dma_heap: if you get the error
kind "Error: open /dev/dma_heap: permission denied", there's a known issue about it (directory mislabelled with selinux). Workaround:
sudo setenforce 0
sudo chcon system\_u:object\_r:device\_t:s0 /dev/dma\_heap/
sudo setenforce 1- If you see errors related to go, you may not have go
$PATHconfigured as root. Make sure it is configured, or define it while runningkind.sh:
sudo PATH=$PATH:/usr/local/go/bin ./kind.sh -ep podman-
You can create your own KIND J2 configuration file if the default one is not sufficient
-
You can also specify these values as environment variables. Command line parameters will override the environment variables.
-
To tear down the KIND cluster when finished simply run
$ ./kind.sh --delete
This section describes the configuration needed for IPv6 and dual-stack environments.
For KIND clusters using KIND v0.7.0 or older (CI currently is using v0.8.1), to use IPv6, IPv6 needs to be enable in Docker on the host:
$ sudo vi /etc/docker/daemon.json
{
"ipv6": true
}
$ sudo systemctl reload docker
On a CentOS host running Docker version 19.03.6, the above configuration worked.
After the host was rebooted, Docker failed to start. To fix, change
daemon.json as follows:
$ sudo vi /etc/docker/daemon.json
{
"ipv6": true,
"fixed-cidr-v6": "2001:db8:1::/64"
}
$ sudo systemctl reload docker
IPv6 from Docker repo provided the fix. Newer documentation does not include this change, so change may be dependent on Docker version.
To verify IPv6 is enabled in Docker, run:
$ docker run --rm busybox ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
inet6 ::1/128 scope host
valid_lft forever preferred_lft forever
341: eth0@if342: <BROADCAST,MULTICAST,UP,LOWER_UP,M-DOWN> mtu 1500 qdisc noqueue
link/ether 02:42:ac:11:00:02 brd ff:ff:ff:ff:ff:ff
inet 172.17.0.2/16 brd 172.17.255.255 scope global eth0
valid_lft forever preferred_lft forever
inet6 2001:db8:1::242:ac11:2/64 scope global flags 02
valid_lft forever preferred_lft forever
inet6 fe80::42:acff:fe11:2/64 scope link tentative
valid_lft forever preferred_lft forever
For the eth0 vEth-pair, there should be the two IPv6 entries (global and link addresses).
Currently, to run OVN-Kubernetes with IPv6 only in a KIND deployment, firewalld needs to be disabled. To disable:
sudo systemctl stop firewalld
NOTE: To run with IPv4, firewalld needs to be enabled, so to reenable:
sudo systemctl start firewalld
If firewalld is enabled during a IPv6 deployment, additional nodes fail to join the cluster:
:
Creating cluster "ovn" ...
✓ Ensuring node image (kindest/node:v1.18.2) 🖼
✓ Preparing nodes 📦 📦 📦
✓ Writing configuration 📜
✓ Starting control-plane 🕹️
✓ Installing StorageClass 💾
✗ Joining worker nodes 🚜
ERROR: failed to create cluster: failed to join node with kubeadm: command "docker exec --privileged ovn-worker kubeadm join --config /kind/kubeadm.conf --ignore-preflight-errors=all --v=6" failed with error: exit status 1
And logs show:
I0430 16:40:44.590181 579 token.go:215] [discovery] Failed to request cluster-info, will try again: Get https://[2001:db8:1::242:ac11:3]:6443/api/v1/namespaces/kube-public/configmaps/cluster-info?timeout=10s: dial tcp [2001:db8:1::242:ac11:3]:6443: connect: permission denied
Get https://[2001:db8:1::242:ac11:3]:6443/api/v1/namespaces/kube-public/configmaps/cluster-info?timeout=10s: dial tcp [2001:db8:1::242:ac11:3]:6443: connect: permission denied
This issue was reported upstream in KIND 1257 and blamed on firewalld.
To run OVN-Kubernetes with IPv6 in a KIND deployment, run:
$ go get github.com/ovn-org/ovn-kubernetes; cd $GOPATH/src/github.com/ovn-org/ovn-kubernetes
$ cd go-controller/
$ make
$ cd ../dist/images/
$ make fedora
$ cd ../../contrib/
$ KIND_IPV4_SUPPORT=false KIND_IPV6_SUPPORT=true ./kind.sh
Once kind.sh completes, setup kube config file:
$ cp ~/ovn.conf ~/.kube/config
-- OR --
$ KUBECONFIG=~/ovn.conf
Once testing is complete, to tear down the KIND deployment:
$ kind delete cluster --name ovn
Currently, IP dual-stack is not fully supported in:
- Kubernetes
- KIND
- OVN-Kubernetes
Kubernetes has some IP dual-stack support but the feature is not complete.
Additional changes are constantly being added. This setup is using the latest
Kubernetes release to test against. Kubernetes is being installed below using
OVN-Kubernetes KIND script, however to test, an equivalent version of kubectl
needs to be installed.
First determine what version of kubectl is currently being used and save it:
$ which kubectl
/usr/bin/kubectl
$ kubectl version --client
Client Version: version.Info{Major:"1", Minor:"17", GitVersion:"v1.17.3", GitCommit:"06ad960bfd03b39c8310aaf92d1e7c12ce618213", GitTreeState:"clean", BuildDate:"2020-02-11T18:14:22Z", GoVersion:"go1.13.6", Compiler:"gc", Platform:"linux/amd64"}
sudo mv /usr/bin/kubectl /usr/bin/kubectl-v1.17.3
sudo ln -s /usr/bin/kubectl-v1.17.3 /usr/bin/kubectl
Download and install latest version of kubectl:
$ K8S_VERSION=v1.24.0
$ curl -LO https://storage.googleapis.com/kubernetes-release/release/$K8S_VERSION/bin/linux/amd64/kubectl
$ chmod +x kubectl
$ sudo mv kubectl /usr/bin/kubectl-v1.18.0
$ sudo rm /usr/bin/kubectl
$ sudo ln -s /usr/bin/kubectl-v1.18.0 /usr/bin/kubectl
$ kubectl version --client
Client Version: version.Info{Major:"1", Minor:"18", GitVersion:"v1.18.0", GitCommit:"9e991415386e4cf155a24b1da15becaa390438d8", GitTreeState:"clean", BuildDate:"2020-03-25T14:58:59Z", GoVersion:"go1.13.8", Compiler:"gc", Platform:"linux/amd64"}
For dual-stack, IPv6 needs to be enable in Docker on the host same as for IPv6 only. See above: Docker Changes For IPv6
IP dual-stack is not currently supported in KIND. There is a PR (692) with IP dual-stack changes. Currently using this to test with.
Optionally, save previous version of KIND (if it exists):
cp $GOPATH/bin/kind $GOPATH/bin/kind.orig
To build locally (if additional needed):
go get github.com/kubernetes-sigs/kind; cd $GOPATH/src/github.com/kubernetes-sigs/kind
git pull --no-edit --strategy=ours origin pull/692/head
make clean
make install INSTALL_DIR=$GOPATH/bin
For status of IP dual-stack in OVN-Kubernetes, see 1142.
To run OVN-Kubernetes with IP dual-stack in a KIND deployment, run:
$ go get github.com/ovn-org/ovn-kubernetes; cd $GOPATH/src/github.com/ovn-org/ovn-kubernetes
$ cd go-controller/
$ make
$ cd ../dist/images/
$ make fedora
$ cd ../../contrib/
$ KIND_IPV4_SUPPORT=true KIND_IPV6_SUPPORT=true K8S_VERSION=v1.24.0 ./kind.sh
Once kind.sh completes, setup kube config file:
$ cp ~/ovn.conf ~/.kube/config
-- OR --
$ KUBECONFIG=~/ovn.conf
Once testing is complete, to tear down the KIND deployment:
$ kind delete cluster --name ovn
In order to use an image/tag other than the default hardcoded in kind.sh, specify one (or both of) the following variables:
$ cd ../../contrib/
$ KIND_IMAGE=example.com/kindest/node K8S_VERSION=v1.24.0 ./kind.sh
This is subject to change because code is being updated constantly. But this is more a cautionary note that this feature is not completely working at the moment.
The nodes do not go to ready because the OVN-Kubernetes hasn't setup the network completely:
$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
ovn-control-plane NotReady master 94s v1.18.0
ovn-worker NotReady <none> 61s v1.18.0
ovn-worker2 NotReady <none> 62s v1.18.0
$ kubectl get pods -o wide --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE
kube-system coredns-66bff467f8-hh4c9 0/1 Pending 0 2m45s <none> <none>
kube-system coredns-66bff467f8-vwbcj 0/1 Pending 0 2m45s <none> <none>
kube-system etcd-ovn-control-plane 1/1 Running 0 2m56s 172.17.0.2 ovn-control-plane
kube-system kube-apiserver-ovn-control-plane 1/1 Running 0 2m56s 172.17.0.2 ovn-control-plane
kube-system kube-controller-manager-ovn-control-plane 1/1 Running 0 2m56s 172.17.0.2 ovn-control-plane
kube-system kube-scheduler-ovn-control-plane 1/1 Running 0 2m56s 172.17.0.2 ovn-control-plane
local-path-storage local-path-provisioner-774f7f8fdb-msmd2 0/1 Pending 0 2m45s <none> <none>
ovn-kubernetes ovnkube-db-cf4cc89b7-8d4xq 2/2 Running 0 107s 172.17.0.2 ovn-control-plane
ovn-kubernetes ovnkube-master-87fb56d6d-7qmnb 3/3 Running 0 107s 172.17.0.2 ovn-control-plane
ovn-kubernetes ovnkube-node-278l9 2/3 Running 0 107s 172.17.0.3 ovn-worker2
ovn-kubernetes ovnkube-node-bm7v6 2/3 Running 0 107s 172.17.0.2 ovn-control-plane
ovn-kubernetes ovnkube-node-p4k4t 2/3 Running 0 107s 172.17.0.4 ovn-worker
Some environments (Fedora32,31 on desktop), have problems when the cluster
is deleted directly with kind kind delete cluster --name ovn, it restarts the host.
The root cause is unknown, this also can not be reproduced in Ubuntu 20.04 or
with Fedora32 Cloud, but it does not happen if we clean first the ovn-kubernetes resources.
You can use the following command to delete the cluster:
contrib/kind.sh --delete