ZENKO-887 change to Installation/Install_guide.rst #240
Conversation
|
|
||
| Helm can now install applications on the Kubernetes cluster. | ||
|
|
||
| 3. Declare the ZooKeeper repository: |
There was a problem hiding this comment.
This section is no longer needed
| $ helm repo add zenko-zookeeper https://scality.github.io/zenko-zookeeper/charts | ||
| "zenko-zookeeper" has been added to your repositories | ||
|
|
||
| $ helm repo add incubator http://storage.googleapis.com/kubernetes-charts-incubator |
There was a problem hiding this comment.
While these are no longer needed we need the following step helm repo add scality https://scality.github.io/charts/
Fixed a couple of formatting errors, one typo. Removed ZENKO-633 Fixed more minor formatting issues. updated issues per Salim's comments removed: autoscaling lines 206-208 added note re NGINX ingress controller Implementing Giacomo's changes. Also added info for non-MetalK8s Helm installation. picked a couple of nits, added nodes 4 and 5 to cluster build intructions.
wabernat
force-pushed
the
bugfix/ZENKO-887_minimum_rootFS_req
branch
from
b5952c7
to
55a20a2
Compare
giacomoguiulfo
changed the title
| Cluster," below. | ||
|
|
||
| The following section describes general cluster requirements, which are tested | ||
| on Metal K8s. Because MetalK8s is designed to operate without support from |
| VMs) running CentOS_ 7.4 (The recommended mimimum for Zenko production service | ||
| is five server nodes with three masters/etcds, but for testing and | ||
| familiarization, three masters and three nodes is fine). You must have SSH | ||
| access to these machines and they must have SSH access to each other. (You |
There was a problem hiding this comment.
No need for SSH access between nodes?
There was a problem hiding this comment.
I don't understand the distinction you're making.
There was a problem hiding this comment.
they must have SSH access to each other
That's not a requirement, and if it were it shouldn't be one.
There was a problem hiding this comment.
OK, I talked through this. Fixed.
| is five server nodes with three masters/etcds, but for testing and | ||
| familiarization, three masters and three nodes is fine). You must have SSH | ||
| access to these machines and they must have SSH access to each other. (You | ||
| can copy SSH credentials from one machine to the next and log in once to |
There was a problem hiding this comment.
Never copy SSH private keys to other machines... There's no need to do it when deploying MetalK8s, and it's bad practice. Worst case you use agent forwarding.
| Sizing | ||
| ====== | ||
|
|
||
| Sizing for Metal K8s |
| - Storage (for the system) | ||
|
|
||
| - 20 GB for the root filesystem | ||
| - 16 GB for etcd |
There was a problem hiding this comment.
If we size etcd storage, we should document how to set up the system(s) to have etcd storage not on / (which is good practice in any case).
There was a problem hiding this comment.
To whose expertise shall I appeal for this information?
There was a problem hiding this comment.
Is this a show-stopper? I need to know either: a) that I can omit the etcd sizing requirement or b) how to configure the system to put etcd somewhere other than / .
If I don't hear back, I'm filing this under "nice to have" and moving the other changes along.
There was a problem hiding this comment.
I think this ought to be discussed with out TS folks. Right now, we don't have any recommendations for putting various types of data on different volumes, but maybe that should be the case.
| To install MetalK8s, you must issue commands from within a virtual shell. | ||
| The following steps ensure you can access the virtual environment. | ||
|
|
||
| 1. Install python-virtualenv: |
There was a problem hiding this comment.
No need to install python-virtualenv?
|
|
||
| 1. Make sure kubectl is installed on your local machine:: | ||
|
|
||
| $ yum install kubectl |
There was a problem hiding this comment.
You get this in a make shell environment, which also ensures you get a version which works with the deployed cluster, which is definitely not guaranteed when installing from CentOS repositories.
| +-------------------------+---------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------------------------------+ | ||
| | `Grafana`_ | Monitoring dashboards for cluster services | http://localhost:8001/api/v1/namespaces/kube-ops/services/kube-prometheus-grafana:http/proxy/ | | | ||
| +-------------------------+---------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------------------------------+ | ||
| | `Cerebro`_ | An administration and monitoring console for | http://localhost:8001/api/v1/namespaces/kube-ops/services/cerebro:http/proxy/ | When accessing Cerebro, connect it to | |
There was a problem hiding this comment.
We pre-provision a config.
There was a problem hiding this comment.
How would you have this reflected in the documentation? Is Cerebro preconfigured to elasticsearch:9200?
There was a problem hiding this comment.
Following precedent in ZENKO-921 (Kibana, next comment) and deleting setup note.
There was a problem hiding this comment.
Yeah, there's an entry in a pre-populated list of 'known clusters' when you open Cerebro.
| | | Elasticsearch clusters | | http://elasticsearch:9200 to operate | | ||
| | | | | the MetalK8s Elasticsearch cluster. | | ||
| +-------------------------+---------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------------------------------+ | ||
| | `Kibana`_ | A search console for logs indexed in Elasticsearch | http://localhost:8001/api/v1/namespaces/kube-ops/services/http:kibana:/proxy/ | When accessing Kibana for the first | |
There was a problem hiding this comment.
We pre-provision a config.
| | | | | field name*. | | ||
| +-------------------------+---------------------------------------------------------+-------------------------------------------------------------------------------------------------+---------------------------------------+ | ||
|
|
||
| See :doc:`../architecture/cluster-services` for more about these services |
There was a problem hiding this comment.
Removed bad docs link.
Can't comment about build other than to note that the merge request went through, and the linter for RST is quite permissive/nonexistent.
There was a problem hiding this comment.
This would only be found by actually building the docs. For MetalK8s, we're working on a step in CI which performs this, and as such makes sure any proposed change to the docs doesn't break its build.
|
Folks, this commit (although merged) contains information that is not pertinent to Zenko IMO. How to install a MetalK8s cluster should be a link to MetalK8s docs. The instructions to install Zenko are also too MetalK8s-centric which is limiting. I think we should revert this change. And before going any further, I would suggest to focus on agreeing on a common architecture for the whole Zenko documentation, setup the infrastructure to build it properly (take MetalK8s docs as an example to follow) and put the content we have in various formats/sources into one coherent information architecture. Such conversation already started on #149 |
|
@smaffulli I think the reason why we were okay with having this MetalK8s info on Zenko was because so many devs were having issues with specifically installing Zenko on MetalK8s. The docs for MetalK8s is great for just installing MetalK8s but we need a bit more configurations for Zenko on MetalK8s (mainly volume provisioning) And if our devs were having pretty much the same issues with installs, likely others will run into the same issues as well. |
This makes a one-line change to Install_guide.rst. Note, however, that I had to git add this file and directory to and paste in a copy of the file because the previous two versions are hung up in builds, so the history may look odd.