Warning: Draft Content
This wiki is under construction - this means that content here may be not fully specified or missing.
TODO: determine/fix containers not ready, get DCAE yamls working, fix health tracking issues for healing
The OOM (ONAP Operation Manager) project has pushed Kubernetes based deployment code to the oom repository. This page details getting ONAP running (specifically the vFirewall demo) on Kubernetes for various virtual and native environments.
Undercloud Installation
We need a kubernetes installation either a base installation or with a thin API wrapper like Rancher or Redhat
There are several options - currently Rancher is a focus as a thin wrapper on Kubernetes.
OS | VIM | Description | Status | Nodes | Links |
---|---|---|---|---|---|
Ubuntu 16.04.2
| Bare Metal VMWare | Rancher | Recommended approach Issue with kubernetes support only in 1.12 (obsolete docker-machine) on OSX | 1-4 | http://rancher.com/docs/rancher/v1.6/en/quick-start-guide/ |
AWS EC2 | EC2 VM's (not ECS or EBS PaaS) | Interesting - in the queue... | x | ||
Linux | Bare Metal or VM | Kubernetes Directly on Ubuntu 16 (no Rancher) | In progress | 1 | https://kubernetes.io/docs/setup/independent/install-kubeadm/ https://lukemarsden.github.io/docs/getting-started-guides/kubeadm/ |
OSX Linux | CoreOS | On Vagrant (Thanks Yves) | Issue: the coreos VM 19G size is insufficient | 1 | https://coreos.com/kubernetes/docs/latest/kubernetes-on-vagrant-single.html Implement OSX fix for Vagrant 1.9.6 https://github.com/mitchellh/vagrant/issues/7747 Avoid the kubectl lock https://github.com/coreos/coreos-kubernetes/issues/886 |
OSX | MInikube on VMWare Fusion VM | minikube VM not restartable | 1 | https://github.com/kubernetes/minikube | |
RHEL 7.3 | VMWare VM | Redhat Kubernetes | services deploy, fix kubectl exec | 1 | https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux_atomic_host/7/html-single/getting_started_with_kubernetes/ |
ONAP Installation
Quickstart Installation
ONAP deployment in kubernetes is modelled in the oom project as a 1:1 set of service:pod sets (1 pod per docker container). The fastest way to get ONAP Kubernetes up is via Rancher.
Primary platform is virtual Ubuntu 16.04 VMs on VMWare Workstation 12.5 on a up to two separate 64Gb/6-core 5820K Windows 10 systems.
Secondary platform is bare-metal 4 NUCs (i7/i5/i3 with 16G each)
Install only the 1.12.x (currently 1.12.6) version of Docker (the only version that works with Kubernetes in Rancher 1.6) Install rancher (use 8880 instead of 8080) In Rancher UI (http://127.0.0.1:8880) , Set IP name of master node in config, create a new onap environment as Kubernetes (will setup kube containers), stop default environment register your host(s) - run following on each host (get from "add host" menu) - install docker 1.12 if not already on the host curl https://releases.rancher.com/install-docker/1.12.sh | sh install kubectl paste kubectl config from rancher mkdir ~/.kube vi ~/.kube/config clone oom (scp your onap_rsa private key first) git clone ssh://michaelobrien@gerrit.onap.org:29418/oom fix nexus3 security temporarily for - OOM-3Getting issue details... STATUS vi oom/kubernetes/oneclick/createAll.bash create_namespace() { kubectl create namespace $1-$2 + kubectl --namespace $1-$2 create secret docker-registry regsecret --docker-server=nexus3.onap.org:10001 --docker-username=docker --docker-password=docker --docker-email=email@email.com + kubectl --namespace $1-$2 patch serviceaccount default -p '{"imagePullSecrets": [{"name": "regsecret"}]}' } Wait until all the hosts show green in rancher, then run the script that wrapps all the kubectl commands run the one time config pod (with mounts for all the other pods) - the pod will stop normally cd oom/kubernetes/config root@obriensystemsucont0:~/onap/oom/kubernetes/config# kubectl create -f pod-config-init.yaml pod "config-init" created Fix DNS resolution before running any more pods ( add service.ns.svc.cluster.local root@obriensystemskub0:~/oom/kubernetes/oneclick# cat /etc/resolv.conf nameserver 192.168.241.2 search localdomain service.ns.svc.cluster.local cd ../oneclick Wait until the containers are all up - you should see... |
Four host Kubernetes cluster in Rancher
In this case 4 Intel NUCs running Ubuntu 16.04.2 natively
Target Deployment State
root@obriensystemsucont0:~/onap/oom/kubernetes/oneclick# kubectl get pods --all-namespaces -o wide
issue may be the fact we need to initialize mounts for pods in the config pod pod-config-init.yaml - rerunning stack
below any colored container has issues getting to running state (currently 10 of 33)
NAMESPACE master:20170705 | NAME | READY | STATUS | RESTARTS (in 14h) | Host | Notes |
---|---|---|---|---|---|---|
onap-aai | aai-service-346921785-624ss | 1/1 | Running | 0 | 1 | |
onap-aai | hbase-139474849-7fg0s | 1/1 | Running | 0 | 2 | |
onap-aai | model-loader-service-1795708961-wg19w | 0/1 | Init:1/2 | 82 | 2 | |
onap-appc | appc-2044062043-bx6tc | 1/1 | Running | 0 | 1 | |
onap-appc | appc-dbhost-2039492951-jslts | 1/1 | Running | 0 | 2 | |
onap-appc | appc-dgbuilder-2934720673-mcp7c | 1/1 | Running | 0 | 2 | |
onap-dcae | not yet pushed | Note: currently there are no DCAE containers running yet (we are missing 6 yaml files (1 for the controller and 5 for the collector,staging,3-cdap pods)) - therefore DMaaP, VES collectors and APPC actions as the result of policy actions (closed loop) - will not function yet. | ||||
onap-dcae-cdap | not yet pushed | |||||
onap-dcae-stg | not yet pushed | |||||
onap-dcae-coll | not yet pushed | |||||
onap-message-router | dmaap-3842712241-gtdkp | 0/1 | CrashLoopBackOff | 164 | 1 | |
onap-message-router | global-kafka-89365896-5fnq9 | 1/1 | Running | 0 | 2 | |
onap-message-router | zookeeper-1406540368-jdscq | 1/1 | Running | 0 | 1 | |
onap-mso | mariadb-2638235337-758zr | 1/1 | Running | 0 | 1 | |
onap-mso | mso-3192832250-fq6pn | 0/1 | CrashLoopBackOff | 167 | 2 | fixed by stopped config-init pod |
onap-policy | brmsgw-568914601-d5z71 | 0/1 | Init:0/1 | 82 | 1 | fixed by stopped config-init pod |
onap-policy | drools-1450928085-099m2 | 0/1 | Init:0/1 | 82 | 1 | |
onap-policy | mariadb-2932363958-0l05g | 1/1 | Running | 0 | 0 | |
onap-policy | nexus-871440171-tqq4z | 0/1 | Running | 0 | 2 | |
onap-policy | pap-2218784661-xlj0n | 1/1 | Running | 0 | 1 | |
onap-policy | pdp-1677094700-75wpj | 0/1 | Init:0/1 | 82 | 2 | fixed by stopped config-init pod |
onap-policy | pypdp-3209460526-bwm6b | 0/1 | Init:0/1 | 82 | 2 | fixed by stopped config-init pod |
onap-portal | portalapps-1708810953-trz47 | 0/1 | Init:CrashLoopBackOff | 163 | 2 | Initial dockerhub mariadb download issue - fixed |
onap-portal | portaldb-3652211058-vsg8r | 1/1 | Running | 0 | 0 | |
onap-portal | vnc-portal-948446550-76kj7 | 0/1 | Init:0/5 | 82 | 1 | |
onap-robot | robot-964706867-czr05 | 1/1 | Running | 0 | 2 | |
onap-sdc | sdc-be-2426613560-jv8sk | 0/1 | Init:0/2 | 82 | 2 | |
onap-sdc | sdc-cs-2080334320-95dq8 | 0/1 | CrashLoopBackOff | 163 | 2 | fixed by stopped config-iiti pod |
onap-sdc | sdc-es-3272676451-skf7z | 1/1 | Running | 0 | 1 | |
onap-sdc | sdc-fe-931927019-nt94t | 0/1 | Init:0/1 | 82 | 1 | |
onap-sdc | sdc-kb-3337231379-8m8wx | 0/1 | Init:0/1 | 82 | 1 | fixed by stopped config-init pod |
onap-sdnc | sdnc-1788655913-vvxlj | 1/1 | Running | 0 | 0 | |
onap-sdnc | sdnc-dbhost-240465348-kv8vf | 1/1 | Running | 0 | 0 | |
onap-sdnc | sdnc-dgbuilder-4164493163-cp6rx | 1/1 | Running | 0 | 0 | |
onap-sdnc | sdnc-portal-2324831407-50811 | 0/1 | Running | 25=vm 0=nuc | 1 | |
onap-vid | vid-mariadb-4268497828-81hm0 | 0/1 | CrashLoopBackOff | 169 | 2 | fixed by config-init and resolv.conf |
onap-vid | vid-server-2331936551-6gxsp | 0/1 | Init:0/1 | 82 | 1 |
I get the same container issues on 3 different deployments (virtual Ubuntu on 2 separate VMWare based machines, and a 3 node NUC cluster). For example the APPC 3-pod service is running fine.
But the 2-pod VID service has failed startup.
Later VID is fixed by running the config-init pod first and also adding to resolv.conf
Cloning details
Install the latest version of the OOM (ONAP Operations Manager) project repo - specifically the ONAP on Kubernetes work just uploaded June 2017
https://gerrit.onap.org/r/gitweb?p=oom.git
git clone ssh://yourgerrituserid@gerrit.onap.org:29418/oom cd oom/kubernetes/oneclick Versions oom : master (1.1.0-SNAPSHOT) onap deployments: 1.0.0 |
---|
Rancher environment for Kubernetes
Adding hosts to the Kubernetes environment will kick in k8s containers
Rancher kubectl config
Nexus3 security settings
Fix nexus3 security for each namespace
in createAll.bash add the following two lines just before namespace creation - to create a secret and attach it to the namespace (thanks to Jason Hunt of IBM last friday to helping us attach it - when we were all getting our pods to come up). A better fix for the future will be to pass these in as parameters from a prod/stage/dev ecosystem config.
create_namespace() { kubectl create namespace $1-$2 + kubectl --namespace $1-$2 create secret docker-registry regsecret --docker-server=nexus3.onap.org:10001 --docker-username=docker --docker-password=docker --docker-email=email@email.com + kubectl --namespace $1-$2 patch serviceaccount default -p '{"imagePullSecrets": [{"name": "regsecret"}]}' } |
---|
Monitor Container Deployment
first verify your kubernetes system is up
Kubernetes specific config
https://kubernetes.io/docs/user-guide/kubectl-cheatsheet/
Dashboard
start the dashboard at http://localhost:8001/ui
kubectl proxy & |
---|
Nexus Docker repo Credentials
Checking out use of a kubectl secret in the yaml files via - https://kubernetes.io/docs/tasks/configure-pod-container/pull-image-private-registry/
Container Endpoint access
Check the services view in the Kuberntes API under robot
robot.onap-robot:88 TCP
robot.onap-robot:30209 TCP
Container Logs
root@obriensystemskub0:~/oom/kubernetes/oneclick# kubectl get pods --all-namespaces -o wide NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE onap-robot robot-44708506-dgv8j 1/1 Running 0 36m 10.42.240.80 obriensystemskub0 root@obriensystemskub0:~/oom/kubernetes/oneclick# kubectl --namespace onap-robot logs -f robot-44708506-dgv8j 2017-07-16 01:55:54: (log.c.164) server started |
---|
SSH into ONAP containers
Normally I would via https://kubernetes.io/docs/tasks/debug-application-cluster/get-shell-running-container/
kubectl exec -it robot -- /bin/bash |
---|
The pod id should be sufficient
root@obriensystemsucont0:~/onap/oom/kubernetes/oneclick# kubectl describe node obriensystemsucont0 | grep robot Namespace Name CPU Requests CPU Limits Memory Requests Memory Limits --------- ---- ------------ ---------- --------------- ------------- |
---|
https://jira.onap.org/browse/OOM-47
in queue....
Running ONAP Portal UI Operations
see Installing and Running the ONAP Demos
In queue.....
Kubernetes Installation Options
Rancher on Ubuntu 16.04
Install Rancher
http://rancher.com/docs/rancher/v1.6/en/quick-start-guide/
http://rancher.com/docs/rancher/v1.6/en/installing-rancher/installing-server/#single-container
Install a docker version that Rancher and Kubernetes support which is currently 1.12.6
http://rancher.com/docs/rancher/v1.5/en/hosts/#supported-docker-versions
curl https://releases.rancher.com/install-docker/1.12.sh | sh |
---|
Verify your Rancher admin console is up on the external port you configured above
Wait for the docker container to finish DB startup
http://rancher.com/docs/rancher/v1.6/en/hosts/
Registering Hosts in Rancher
Having issues registering a combined single VM (controller + host) - use your real IP not localhost
In settings | Host Configuration | set your IP [root@obrien-b2 etcd]# sudo docker run -e CATTLE_AGENT_IP="192.168.163.128" --rm --privileged -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/rancher:/var/lib/rancher rancher/agent:v1.2.2 http://192.168.163.128:8080/v1/scripts/A9487FC88388CC31FB76:1483142400000:IypSDQCtA4SwkRnthKqH53Vxoo |
---|
See your host registered
Troubleshooting
Rancher fails to restart on server reboot
Having issues after a reboot of a colocated server/agent
Docker Nexus Config
- OOM-3Getting issue details... STATUS
Out of the box we cant pull images - currently working on a config step around https://kubernetes.io/docs/tasks/configure-pod-container/pull-image-private-registry/
kubectl create secret docker-registry regsecret --docker-server=nexus3.onap.org:10001 --docker-username=docker --docker-password=docker --docker-email=someone@amdocs.com |
---|
imagePullSecrets: - name: regsecret |
---|
Failed to pull image "nexus3.onap.org:10001/openecomp/testsuite:1.0-STAGING-latest": image pull failed for nexus3.onap.org:10001/openecomp/testsuite:1.0-STAGING-latest, this may be because there are no credentials on this request. details: (unauthorized: authentication required)
kubelet 172.17.4.99
OOM Repo changes
20170629: fix on 20170626 on a hardcoded proxy - (for those who run outside the firewall) - https://gerrit.onap.org/r/gitweb?p=oom.git;a=commitdiff;h=131c2a42541fb807f395fe1f39a8482a53f92c60
DNS resolution
add "service.ns.svc.cluster.local" to fix
Search Line limits were exceeded, some dns names have been omitted, the applied search line is: default.svc.cluster.local svc.cluster.local cluster.local kubelet.kubernetes.rancher.internal kubernetes.rancher.internal rancher.internal
https://github.com/rancher/rancher/issues/9303
root@obriensystemskub0:~/oom/kubernetes/oneclick# cat /etc/resolv.conf
# Dynamic resolv.conf(5) file for glibc resolver(3) generated by resolvconf(8)
# DO NOT EDIT THIS FILE BY HAND -- YOUR CHANGES WILL BE OVERWRITTEN
nameserver 192.168.241.2
search localdomain service.ns.svc.cluster.local
Deprecated Kubernetes Installation Options
Bare RHEL 7.3 VM - Multi Node Cluster
In progress as of 20170701
https://kubernetes.io/docs/getting-started-guides/scratch/
https://github.com/kubernetes/kubernetes/releases/latest https://github.com/kubernetes/kubernetes/releases/tag/v1.7.0 https://github.com/kubernetes/kubernetes/releases/download/v1.7.0/kubernetes.tar.gz tar -xvf kubernetes.tar optional build from source vi Vagrantfile go directly to binaries /run/media/root/sec/onap_kub/kubernetes/cluster ./get-kube-binaries.sh export Path=/run/media/root/sec/onap_kub/kubernetes/client/bin:$PATH [root@obrien-b2 server]# pwd /run/media/root/sec/onap_kub/kubernetes/server kubernetes-manifests.tar.gz kubernetes-salt.tar.gz kubernetes-server-linux-amd64.tar.gz README tar -xvf kubernetes-server-linux-amd64.tar.gz /run/media/root/sec/onap_kub/kubernetes/server/kubernetes/server/bin build images [root@obrien-b2 etcd]# make
(go lang required - adjust google docs) https://golang.org/doc/install?download=go1.8.3.linux-amd64.tar.gz |
---|
CoreOS on Vagrant on RHEL/OSX
(Yves alerted me to this) - currently blocked by the 19g VM size (changing the HD of the VM is unsupported in the VirtualBox driver)
https://coreos.com/kubernetes/docs/latest/kubernetes-on-vagrant-single.html
Implement OSX fix for Vagrant 1.9.6 https://github.com/mitchellh/vagrant/issues/7747
Adjust the VagrantFile for your system
NODE_VCPUS = 1 NODE_MEMORY_SIZE = 2048 to (for a 5820K on 64G for example) NODE_VCPUS = 8 NODE_MEMORY_SIZE = 32768 |
---|
curl -O https://storage.googleapis.com/kubernetes-release/release/v1.6.1/bin/darwin/amd64/kubectl chmod +x kubectl skipped (mv kubectl /usr/local/bin/kubectl) - already there ls /usr/local/bin/kubectl git clone https://github.com/coreos/coreos-kubernetes.git cd coreos-kubernetes/single-node/ vagrant box update sudo ln -sf /usr/local/bin/openssl /opt/vagrant/embedded/bin/openssl vagrant up Wait at least 5 min (Yves is good) (rerun from here) export KUBECONFIG="${KUBECONFIG}:$(pwd)/kubeconfig" kubectl config use-context vagrant-single obrienbiometrics:single-node michaelobrien$ export KUBECONFIG="${KUBECONFIG}:$(pwd)/kubeconfig" obrienbiometrics:single-node michaelobrien$ kubectl config use-context vagrant-single Switched to context "vagrant-single". obrienbiometrics:single-node michaelobrien$ kubectl proxy & [1] 4079 obrienbiometrics:single-node michaelobrien$ Starting to serve on 127.0.0.1:8001 goto $ kubectl get nodes $ kubectl get service --all-namespaces $ kubectl cluster-info git clone ssh://michaelobrien@gerrit.onap.org:29418/oom cd oom/kubernetes/oneclick/ obrienbiometrics:oneclick michaelobrien$ ./createAll.bash -n onap **** Done ****obrienbiometrics:oneclick michaelobrien$ kubectl get service --all-namespaces ... onap-vid vid-server 10.3.0.31 <nodes> 8080:30200/TCP 32s obrienbiometrics:oneclick michaelobrien$ kubectl get pods --all-namespaces NAMESPACE NAME READY STATUS RESTARTS AGE kube-system heapster-v1.2.0-4088228293-3k7j1 2/2 Running 2 4h kube-system kube-apiserver-172.17.4.99 1/1 Running 1 4h kube-system kube-controller-manager-172.17.4.99 1/1 Running 1 4h kube-system kube-dns-782804071-jg3nl 4/4 Running 4 4h kube-system kube-dns-autoscaler-2715466192-k45qg 1/1 Running 1 4h kube-system kube-proxy-172.17.4.99 1/1 Running 1 4h kube-system kube-scheduler-172.17.4.99 1/1 Running 1 4h kube-system kubernetes-dashboard-3543765157-qtnnj 1/1 Running 1 4h onap-aai aai-service-346921785-w3r22 0/1 Init:0/1 0 1m ... reset obrienbiometrics:single-node michaelobrien$ rm -rf ~/.vagrant.d/boxes/coreos-alpha/ |
---|
OSX Minikube
curl -LO https://storage.googleapis.com/kubernetes-release/release/$(curl -s https://storage.googleapis.com/kubernetes-release/release/stable.txt)/bin/darwin/amd64/kubectl chmod +x ./kubectl sudo mv ./kubectl /usr/local/bin/kubectl kubectl cluster-info kubectl completion -h brew install bash-completion curl -Lo minikube https://storage.googleapis.com/minikube/releases/v0.19.0/minikube-darwin-amd64 && chmod +x minikube && sudo mv minikube /usr/local/bin/ minikube start --vm-driver=vmwarefusion kubectl run hello-minikube --image=gcr.io/google_containers/echoserver:1.4 --port=8080 kubectl expose deployment hello-minikube --type=NodePort kubectl get pod curl $(minikube service hello-minikube --url) minikube stop |
---|
When upgrading from 0.19 to 0.20 - do a minikube delete
RHEL Kubernetes - Redhat 7.3 Enterprise Linux Host
Running onap kubernetes services in a single VM using Redhat Kubernetes for 7.3
Redhat provides 2 docker containers for the scheduler and nbi components and spins up 2 (# is scalable) pod containers for use by onap.
[root@obrien-mbp oneclick]# docker ps |
---|
Kubernetes setup
Uninstall docker-se (we installed earlier) subscription-manager repos --enable=rhel-7-server-optional-rpms [root@obrien-mbp opt]# ./kubestart.sh [root@obrien-mbp opt]# ss -tulnp | grep -E "(kube)|(etcd)"
|
---|
References
- OOM-1Getting issue details... STATUS