Updates for 3.7 install/upgrade

[adellape]

http://file.rdu.redhat.com/~adellape/112017/installupgrade37/install_config/index.html

[adellape]

[epic] Break out playbooks/roles

[adellape]

Document: Installing new OCP 3.7 clusters

[adellape]

Document: OCP 3.6 -> 3.7 upgrade

[wkshi]

https://bugzilla.redhat.com/show_bug.cgi?id=1509112
https://bugzilla.redhat.com/show_bug.cgi?id=1509137
@sdodson Due to above bugs, quick installer is totally unable to use. Are we still going to release quick installer in doc?

[wkshi]

@adellape
At install_config/install/advanced_install.adoc part, The "Add the hosts listed under [glusterfs] to the [nodes] group as well:" seems give a wrong example ip address:

[glusterfs]
192.168.10.11 glusterfs_ip=192.168.10.11 glusterfs_devices='[ "/dev/xvdc", "/dev/xvdd" ]'
192.168.10.12 glusterfs_ip=192.168.10.12 glusterfs_devices='[ "/dev/xvdc", "/dev/xvdd" ]'
192.168.10.13 glusterfs_ip=192.168.10.13 glusterfs_devices='[ "/dev/xvdc", "/dev/xvdd" ]'

[nodes]
192.168.10.14
192.168.10.15
192.168.10.16

[jianlinliu]

According to playbook order listed in playbooks/common/openshift-cluster/config.yml, seem like there is no single entrypoint to call openshift-prometheus.yml, while it is already included in openshift_hosted.yml, it seem like a duplicate to list openshift-prometheus.yml here, so I suggest we should remove this line here.

[mtnbikenc]

Prometheus does have an entry point [here](https://github.com/openshift/openshift-ansible/blob/release-3.7/playbooks/byo/openshift-cluster/openshift-prometheus.yml. The call to common/openshift-cluster/openshift_prometheus.yml should not be in common/openshift-cluster/openshift_hosted.yml, but should rather be in common/openshift-cluster/config.yml. I noticed this yesterday when doing the refactoring for 3.8. I will put in a PR to update the 3.7 branch accordingly.

[jianlinliu]

Yeah, totally agree to move the call for openshift-prometheus.yml from common/openshift-cluster/openshift_hosted.yml to common/openshift-cluster/config.yml. If that, agree to keep this line here.

[ganhuang]

A couple issues found after having a look of the whole doc:

1. Overriding Detected IP Addresses and Host Names

# ansible-playbook  [-i /path/to/inventory] \
    ~/openshift-ansible/playbooks/byo/config.yml

To get the facts, should run playbook /usr/share/ansible/openshift-ansible/roles/openshift_facts/library/openshift_facts.py

# ansible-playbook  [-i /path/to/inventory] \
    /usr/share/ansible/openshift-ansible/roles/openshift_facts/library/openshift_facts.py

2. Configuring Deployment Type

For a fresh install, we recommend the new parameter openshift_deployment_type instead of deployment_type.

3. Configuring System Containers

You can configure your OpenShift Container Platform installation to run certain components as system containers instead of their RPM or standard containerized methods. Currently, the docker and etcd components can be run as system containers in OpenShift Container Platform.

Besides etcd, docker system containers, we also could enable master, node, openvswitch system containers. Detailed parameters. This is not documented yet.

4. Multiple Masters Examples

Since OCP 3.7, osm_controller_lease_ttl is deprecated, and the new client based leader election code in the controllers is used as the default. Detailed PR

5. Install Methods for Containerized Hosts

When installing an environment with multiple masters, the load balancer cannot be deployed by the installation process as a container. See Advanced Installation for load balancer requirements using the native HA method.

Since OCP 3.6(if my memory is correct), containerized LB has been already supported.

[gpei]

1.http://file.rdu.redhat.com/~adellape/110117/cfme37/install_config/install/rpm_vs_containerized.html#containerized-required-images

The cli_docker_* options were deprecated, updated as below now.
#openshift_docker_additional_registries=registry.example.com
#openshift_docker_insecure_registries=registry.example.com
#openshift_docker_blocked_registries=registry.hacker.com

2. http://file.rdu.redhat.com/~adellape/110117/cfme37/install_config/install/advanced_install.html#advanced-install-custom-certificates

"cafile" key was added in openshift_master_named_certificates option list now

openshift_master_named_certificates=[{"certfile": "/path/to/custom1.crt", "keyfile": "/path/to/custom1.key"}]
was changed to
openshift_master_named_certificates=[{"certfile": "/path/to/custom1.crt", "keyfile": "/path/to/custom1.key", "cafile": "/path/to/custom-ca1.crt"}]

openshift_master_named_certificates=[{"certfile": "/path/to/custom1.crt", "keyfile": "/path/to/custom1.key", "names": ["public-master-host.com"]}]
was changed to
openshift_master_named_certificates=[{"certfile": "/path/to/custom1.crt", "keyfile": "/path/to/custom1.key", "names": ["public-master-host.com"], "cafile": "/path/to/custom-ca1.crt"}]

[adellape]

@ganhuang Thank you! Edits made, except for the following two points (comments inline):

3. Configuring System Containers

You can configure your OpenShift Container Platform installation to run certain components as system containers instead of their RPM or standard containerized methods. Currently, the docker and etcd components can be run as system containers in OpenShift Container Platform.

Besides etcd, docker system containers, we also could enable master, node, openvswitch system containers. Detailed parameters. This is not documented yet.

I will work on this ASAP in a follow-up PR.

5. Install Methods for Containerized Hosts

When installing an environment with multiple masters, the load balancer cannot be deployed by the installation process as a container. See Advanced Installation for load balancer requirements using the native HA method.

Since OCP 3.6(if my memory is correct), containerized LB has been already supported.

I'll need more info on this, so I'll also leave this as is for now and work on it in a follow-up PR.

[adellape]

@wkshi Thanks! Per your comments:

https://bugzilla.redhat.com/show_bug.cgi?id=1509112
https://bugzilla.redhat.com/show_bug.cgi?id=1509137
@sdodson Due to above bugs, quick installer is totally unable to use. Are we still going to release quick installer in doc?

I'm removing the quick installer from the 3.7 docs, and we'll add a release note to explains it. cc @ahardin-rh

At install_config/install/advanced_install.adoc part, The "Add the hosts listed under [glusterfs] to the [nodes] group as well:" seems give a wrong example ip address:

This is now fixed.

[adellape]

@gpei Thanks! Edits made.

[ganhuang]

@adellape

5. Install Methods for Containerized Hosts

When installing an environment with multiple masters, the load balancer cannot be deployed by the installation process as a container. See Advanced Installation for load balancer requirements using the native HA method.

Since OCP 3.6(if my memory is correct), containerized LB has been already supported.
I'll need more info on this, so I'll also leave this as is for now and work on it in a follow-up PR.

This is the bug which was fixed since 3.5

[gpei]

Hi, @adellape another issue found.
In http://file.rdu.redhat.com/~adellape/110117/cfme37/install_config/install/advanced_install.html#advanced-install-etcd-system-container
The data directory for etcd using this method is /var/lib/etcd/etcd.etcd/etcd.etcd/member

The data directory used by etcd system container has been changed to /var/lib/etcd
per openshift/openshift-ansible@657cf2a

[wmengRH]

@adellape
in file architecture/service_catalog/index.adoc, "Service Plan" should be "Cluster Service Plan" for OCP 3.7 GA

apiVersion: servicecatalog.k8s.io/v1beta1
kind: ClusterServicePlan

[adellape]

@ganhuang @gpei @wmengRH Thanks so much! Latest comments addressed via #6517.

[gpei]

@adellape Thanks for the updates! One more typo found in
"Single Master and Multiple Nodes Inventory File" http://file.rdu.redhat.com/~adellape/112817/edit_installupgrade/install_config/install/advanced_install.html#single-master

"etcd " group should be added in [OSEv3:children]

[adellape]

@gpei Thanks! Fixed via #6568. (It was also reported via https://bugzilla.redhat.com/show_bug.cgi?id=1518964).