03ebbaaca3
MAAS uses ipmipower to power nodes on and off. The ipmipower command sometimes fails with the message: x.x.x.x: BMC error but exits with return code of 0. Because MAAS is not aware of this specific error, and is also not able to detect the power state from the output, the built in retry logic is not used, resulting in failed deployments or nodes in Power Error. This patch adds "BMC error" to the list of known retriable errors, and also adds additional retries. Error strings from ipmipower 1.4: http://git.savannah.gnu.org/cgit/freeipmi.git/tree/ipmipower/ipmipower_output.c?h=Release-1_4_0_branch#n52 IPMI_ERRORS known to MAAS 2.3: https://git.launchpad.net/maas/tree/src/provisioningserver/drivers/power/ipmi.py?h=2.3#n50 Change-Id: Ia4b10e47855d67ba81e9ab800be3d780f8b38cac |
||
---|---|---|
.. | ||
maas-rack-controller | ||
maas-region-controller | ||
sstream-cache | ||
README.md |
Ubuntu MaaS Region Controller
Ubuntu MaaS Rack Controller
Overview
The MaaS project attempts to build highly decoupled metal as a service containers for use on the Kubernetes platform. Today, we only break the MaaS service into the traditional region and rack controllers and breaking it down further is a work in progress.
Building Containers
$ make build
Launching on Kubernetes
This will create the bridge necessary for MaaS provisioning (fixed with the name 'maas' rigt now) and launch the region controller and rack controller containers on kubernetes using kubectl by leveraging the YAML manifests in maas/deployments.
$ make kuber_bridge
...
$ make kuber_deploy
sudo kubectl create -f deployment/maas-service.yaml
service "maas-region-ui" created
sudo kubectl create -f deployment/maas-region-deployment.yaml
deployment "maas-region" created
sudo kubectl create -f deployment/maas-rack-deployment.yaml
deployment "maas-rack" created
The provisioning network is fixed (and configured by kuber_bridge) as 10.7.200.0/24. To connect external physical hardware to this network, simply place the network interface into the maas bridge, e.g:
brctl addif maas eth1
To destroy the kubernetes resources, you can run:
$ make kuber_clean
sudo kubectl delete deployment maas-region
deployment "maas-region" deleted
sudo kubectl delete deployment maas-rack
deployment "maas-rack" deleted
sudo kubectl delete service maas-region-ui
service "maas-region-ui" deleted
Once the region controller comes up, and you can login as admin/admin, you must configure a gateway within the UI on the 10.7.200.0 network, setting that to 10.7.200.1. You must also enable DHCP and set the primary rack controller to the maas rack container booted (it will be a drop down choice). This will eventually be automated.
Running Containers
$ make run_region
sudo docker run -d -p 7777:80 -v /sys/fs/cgroup:/sys/fs/cgroup:ro --privileged --name maas-region-controller maas-region:dockerfile
d7462aabf4d8982621c30d7df36adf6c3e0f634701c0a070f7214301829fa92e
$ make run_rack
sudo docker run -d -v /sys/fs/cgroup:/sys/fs/cgroup:ro --privileged --name maas-rack-controller maas-rack:dockerfile
fb36837cd68e56356cad2ad853ae517201ee3349fd1f80039185b71d052c5326
Region Bootstrap
The scripts/create-provision-network.sh
script attempts to bootstrap both an admin user (with the password admin) but also creates a maas provisioning network matching the docker default, namely 172.16.86.0/24. Turning this into a more configurable setting and also allowing for a dedicated provisioning network that can be plugged in via bridging to an actual physical network is a work in progress. However, with the calls we do make you should be able to see the rack controller connected with an active dhcpd process running in the UI.
Retrieving Region Controller Details
Note that retrieving the API key may not be possible as MaaS region initialization is delayed within the containers init startup. It may take 60 seconds or so in order to retrieve the API key, during which you may see the following message:
$ make get_region_api_key
sudo docker exec maas-region-controller maas-region-admin apikey --username maas
WARNING: The maas-region-admin command is deprecated and will be removed in a future version. From now on please use 'maas-region' instead.
CommandError: User does not exist.
make: *** [get_region_api_key] Error 1
When the API is up and the admin user registered you will see the following:
$ make get_region_api_key
sudo docker exec maas-region-controller maas-region apikey --username admin
ksKQbjtTzjZrZy2yP7:jVq2g4x5FYdxDqBQ7P:KGfnURCrYSKmGE6k2SXWk4QVHVSJHBfr
You can also retrieve the region secret and IP address, used to initialize the rack controller:
$ make get_region_secret
sudo docker exec maas-region-controller cat /var/lib/maas/secret && echo
2036ba7575697b03d73353fc72a01686
$ make get_region_ip_address
sudo docker inspect --format '{{ .NetworkSettings.Networks.bridge.IPAddress }}' maas-region-controller
172.16.86.4
Link Rack and Region
Finally, with the output above we can link the region controller with the rack controller by feeding the rack controller the endpoint and secret it requires. Shortly after MaaS will initiate an image sync with the rack.
$ make register_rack -e URL=http://172.16.84.4 SECRET=2036ba7575697b03d73353fc72a01686
sudo docker exec maas-rack-controller maas-rack register --url http://172.16.84.4 --secret 2036ba7575697b03d73353fc72a01686
alan@hpdesktop:~/Workbench/att/airshipit/dockerfiles/maas$
Finally, to access your MaaS UI, visit http://172.0.0.1:7777/MAAS/ and login as admin/admin.