...
Metric | Min | Full System | Notes | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
vCPU | 8 | 64 recommended (16/32 ok) | The full ONAP system of 85+ containers is CPU and Network bound on startup - therefore if you pre pull the docker images to remove the network slowdown - vCPU utilization will peak at 52 cores on a 64 core system and bring the system up in under 4 min. On a system with 8/16 cores you will see the normal 13/7 min startup time as we throttle 52 to 8/16. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RAM | 7g (a couple components) | 55g (75 containers) | Note: you need at least 53g RAM (3g is for Rancher/Kubernetes itself (the VFs exist on a separate openstack). 53 to start and 57 after running the system for a day | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
HD | 60g | 120g+ | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Software | Rancher v1.6.1014 Kubectl 1.8+ .6 (not 1.9 yet) Docker 117.03.122 HELM 2.6+ - tested 2.7.2 8.0 on openlab, Rackspace and AWS - as of master 20180124
|
...
OS | VIM | Description | Status | Nodes | Links | ||
---|---|---|---|---|---|---|---|
Ubuntu 16.04.2
| Bare Metal VMWare | Rancher | Recommended approachIssue with kubernetes support only in | 1.12 (obsolete docker-machine) on OSX | 1-4 | -n | http://rancher.com/docs/rancher/v1.6/en/quick-start-guide/ |
...
***************** Note: uninstall docker if already installed - as Kubernetes only support 1.12.x - as of 201708091.8 under rancher supports 17.03.2 as of 20180124
***************** |
---|
...
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 on any bare metal or VM that supports a clean Ubuntu 16.04 install and more than 50G 55G ram.
TODO: REMOVE from table cell - wrapping is not working
(on each host) add to your /etc/hosts to point your ip to your hostname (add your hostname to the end). Add entries for all other hosts in your cluster.
Open Ports On most hosts like openstack or EC2 you can open all the ports or they are open by default - on some environments like Rackspace VM's you need to open them
Fix virtual memory allocation (to allow onap-log:elasticsearch to come up under Rancher 1.6.11)
clone oom (scp your onap_rsa private key first - or clone anon - Ideally you get a full gerrit account and join the community) see ssh/http/http access links below https://gerrit.onap.org/r/#/admin/projects/oom
or use https (substitute your user/pass)
(on each host (server and client(s) which may be the same machine)) Install only the 117.12.x (currently 1.12.6) 03.2 version of Docker (the only version that works with Kubernetes in Rancher 1.6.13+) Install Docker
Pre pull docker images the first time you install onap. Currently the pre-pull will take 2016-180 min depending on your network. Pre pulling the images will allow the entire ONAP to start in 3-8 min instead of up to 3 hours.
Use script above in oom/kubernetes/config once it is merged. https://git.onap.org/oom/tree/kubernetes/config/prepull_docker.sh
To monitor when prepull is finished see section: Prepulldockerimages. It is advised to wait until pre pull has finished before continuing. (on the master only) Install rancher (Optional: use 8880 instead of 8080 if there is a conflict) - note there may be issues with the dns pod in Rancher after a reboot or when running clustered hosts - a clean system will be OK -
In Rancher UI - dont use (http://127.0.0.1:8880) - use the real IP address - so the client configs are populated correctly with callbacks You must deactivate the default CATTLE environment - by adding a KUBERNETES environment - and Deactivating the older default CATTLE one - your added hosts will attach to the default
Register your hostRegister your host(s) - run following on each host (including the master if you are collocating the master/host on a single machine/vm) For each host, In Rancher > Infrastructure > Hosts. Select "Add Host" The first time you add a host - you will be presented with a screen containing the routable IP - hit save only on a routable IP. Enter IP of host: (if you launched racher with 127.0.0.1/localhost - otherwise keep it empty - it will autopopulate the registration with the real IP) Copy command to register host with Rancher, Execute command on each host, for example:
wait for kubernetes menu to populate with the CLI Install KubectlThe following will install kubectl (1.8 only) https://github.com/kubernetes/kubernetes/issues/57528 on a linux host. Once configured, this client tool will provide management of a Kubernetes cluster.
Paste kubectl config from Rancher (you will see the CLI menu in Rancher / Kubernetes after the k8s pods are up on your host) Click on "Generate Config" to get your content to add into .kube/config Verify that Kubernetes config is good
Install HelmThe following will install Helm (use 2.6+ - tested - currently 2.7.2 so far) 8.0 on a linux host. Helm is used by OOM for package and configuration management. https://lists.onap.org/pipermail/onap-discuss/2018-January/007674.html Prerequisite: Install Kubectl
Undercloud done - move to ONAP Installation You can install OOM manually below or run the cd.sh below or attached to the top of this page - Install/RefreshOOM https://github.com/obrienlabs/onap-root/blob/master/cd.sh manually..... Wait until all the hosts show green in rancher, Then we are ready to configure and deploy onap environment in kubernetes. These scripts are found in the folders:
First source oom/kubernetes/oneclick/setenv.bash. This will set your helm list of components to start/delete
Seconds we need configure the onap before deployment. This is a onetime operation that spawns temporality config pod. This mounts the volume /dockerdata/ contained in the pod config-init and also creates the directory “/dockerdata-nfs” on the kubernetes node. This mount is required for all other ONAP pods to function. Note: the pod will stop after NFS creation - this is normal. https://git.onap.org/oom/tree/kubernetes/config/onap-parameters-sample.yaml
**** Creating configuration for ONAP instance: onap Wait for the config-init pod is gone before trying to bring up a component or all of ONAP - around 60 sec (up to 10 min) - see https://wiki.onap.org/display/DW/ONAP+on+Kubernetes#ONAPonKubernetes-Waitingforconfig-initcontainertofinish-20sec root@ip-172-31-93-122:~/oom_20170908/oom/kubernetes/config# kubectl get pods --all-namespaces -a onap config 0/1 Completed 0 1m Note: When using the -a option the config container will show up with the status, however when not used with the -a flag, it will not be present Cluster Configuration (optional - do not use if your server/client are co-located)3. Share the /dockerdata-nfs Folder between Kubernetes Nodes Running ONAPDon't run all the pods unless you have at least 52G allocated - if you have a laptop/VM with 16G - then you can only run enough pods to fit in around 11G
(to bring up a single service at a time) Use the default "onap" namespace if you want to run robot tests out of the box - as in "onap-robot" Bring up core components
Only if you have >52G run the following (all namespaces)
ONAP is OK if everything is 1/1 or 2/2 in the following
Run the ONAP portal via instructions at RunningONAPusingthevnc-portal Wait until the containers are all up check AAI endpoints root@ip-172-31-93-160:/dockerdata-nfs/onap/robot# kubectl -n onap-aai exec -it aai-service-3321436576-2snd6 bash root@aai-service-3321436576-2snd6:/# ps -ef UID PID PPID C STIME TTY TIME CMD root 1 0 0 15:50 ? 00:00:00 /usr/local/sbin/haproxy-systemd- root 7 1 0 15:50 ? 00:00:00 /usr/local/sbin/haproxy-master root@ip-172-31-93-160:/dockerdata-nfs/onap/robot# curl https://127.0.0.1:30233/aai/v11/service-design-and-creation/models curl: (60) server certificate verification failed. CAfile: /etc/ssl/certs/ca-certificates.crt CRLfile: none |
...