epostnikof/kubespray
1
0
Fork 0
mirror of https://github.com/kubernetes-sigs/kubespray.git synced 2025-12-14 05:45:06 +03:00
Code Issues Packages Projects Releases Wiki Activity

Add HA/LB endpoints for kube-apiserver

Browse Source 
* Add HA docs for API server.
* Add auto-evaluated internal endpoints and clarify the loadbalancer_apiserver
vars and usecases.
* Use facts for kube_apiserver to not repeat code and enable LB endpoints use.
* Use /healthz check for the wait-for apiserver.
* Use the single endpoint for kubelet instead of the list of apiservers
* Specify kube_apiserver_count to for HA layout

Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
This commit is contained in:
Bogdan Dobrelya
2016-07-13 17:13:47 +02:00
parent af4c41f32e
commit 731d32afda
11 changed files with 130 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

92
docs/ha-mode.md
View File

@@ -31,4 +31,94 @@ flannel networking plugin always uses a single `--etcd-server` endpoint!
Kube-apiserver
--------------
TODO(bogdando) TBD
K8s components require a loadbalancer to access the apiservers via a reverse
proxy. A kube-proxy does not support multiple apiservers for the time being so
you will need to configure your own loadbalancer to achieve HA. Note that
deploying a loadbalancer is up to a user and is not covered by ansible roles
in Kargo. By default, it only configures a non-HA endpoint, which points to
the `access_ip` or IP address of the first server node in the `kube-master`
group. It can also configure clients to use endpoints for a given loadbalancer
type.
A loadbalancer (LB) may be an external or internal one. An external LB
provides access for external clients, while the internal LB accepts client
connections only to the localhost, similarly to the etcd-proxy HA endpoints.
Given a frontend `VIP` address and `IP1, IP2` addresses of backends, here is
an example configuration for a HAProxy service acting as an external LB:
```
listen kubernetes-apiserver-https
bind <VIP>:8383
option ssl-hello-chk
mode tcp
timeout client 3h
timeout server 3h
server master1 <IP1>:443
server master2 <IP2>:443
balance roundrobin
```
And the corresponding example global vars config:
```
apiserver_loadbalancer_domain_name: "lb-apiserver.kubernetes.local"
loadbalancer_apiserver:
address: <VIP>
port: 8383
```
This domain name, or default "lb-apiserver.kubernetes.local", will be inserted
into the `/etc/hosts` file of all servers in the `k8s-cluster` group. Note that
the HAProxy service should as well be HA and requires a VIP management, which
is out of scope of this doc.
The internal LB may be the case if you do not want to operate a VIP management
HA stack and require no external and no secure access to the K8s API. The group
var `loadbalancer_apiserver_localhost` (defaults to `false`) controls that
deployment layout. When enabled, it is expected each node in the `k8s-cluster`
group to run a loadbalancer that listens the localhost frontend and has all
of the apiservers as backends. Here is an example configuration for a HAProxy
service acting as an internal LB:
```
listen kubernetes-apiserver-http
bind localhost:8080
mode tcp
timeout client 3h
timeout server 3h
server master1 <IP1>:8080
server master2 <IP2>:8080
balance leastconn
```
And the corresponding example global vars config:
```
loadbalancer_apiserver_localhost: true
```
This var overrides an external LB configuration, if any. Note that for this
example, the `kubernetes-apiserver-http` endpoint has backends receiving
unencrypted traffic, which may be a security issue when interconnecting
different nodes, or may be not, if those belong to the isolated management
network without external access.
In order to achieve HA for HAProxy instances, those must be running on the
each node in the `k8s-cluster` group as well, but require no VIP, thus
no VIP management.
Access endpoints are evaluated automagically, as the following:
| Endpoint type | kube-master | non-master |
|------------------------------|---------------|---------------------|
| Local LB (overrides ext) | http://lc:p | http://lc:p |
| External LB, no internal | https://lb:lp | https://lb:lp |
| No ext/int LB (default) | http://lc:p | https://m[0].aip:sp |
Where:
* `m[0]` - the first node in the `kube-master` group;
* `lb` - LB FQDN, `apiserver_loadbalancer_domain_name`;
* `lc` - localhost;
* `p` - insecure port, `kube_apiserver_insecure_port`
* `sp` - secure port, `kube_apiserver_port`;
* `lp` - LB port, `loadbalancer_apiserver.port`, defers to the secure port;
* `ip` - the node IP, defers to the ansible IP;
* `aip` - `access_ip`, defers to the ip.