RFE/Question - Turn PV examples into templates

[tbielawa]

In openshift/openshift-ansible#5336 I converted the server and the postgres db PVs into templates with parameters to make it easier and more natural to use with the openshift-ansible oc_* tools (process, create, etc).

I create the PV templates in here:

• https://github.com/openshift/openshift-ansible/pull/5336/files#diff-408ea6dc0d53b2060afcab9004f3d1e7

And then later I conditionally create each template in here:

• https://github.com/openshift/openshift-ansible/pull/5336/files#diff-0049ddfb751973936061ec914a7a967eR30
• https://github.com/openshift/openshift-ansible/pull/5336/files#diff-0049ddfb751973936061ec914a7a967eR44 (I realize that has the wrong task name, it should say CFME DB PV, I'll update it)

Here's the server PV template and the DB template file:

Server Template

• https://github.com/openshift/openshift-ansible/pull/5336/files#diff-3d51858af4ddb5e63305e906ceaf8c20

DB Template

• https://github.com/openshift/openshift-ansible/pull/5336/files#diff-cfe1b5ebff1a423bbd197ee2676a4e0c

Would you consider shipping these updated versions in your repo?

They're a little rough right now, so, if you're open to this then I'll submit a PR and we can get them using language/formatting you prefer.

Thanks!

[carbonin]

I'm 👍 on this.

Our PV examples are, well, just examples right now. This would allow someone to create a pv directly using the oc tooling without editing the files (and dirtying their local git status).

The only question I would have is around how we do backup and restore. @fbladilo would this change that process at all? In particular I'm concerned about the hard-coded values for things like the backup pvc, would we also want to template that as a part of this change to be consistent?

[carbonin]

Still would want @bdunne and @fbladilo to comment on this about the general direction. You guys have anything to add?

[tbielawa]

Some more information

What I did with the example templates you provide was fairly limited:

• I changed it to the kube template form, which means the FIRST kind tag in the inventory is with value template, then some labels, annots, etc. Then there's a objects section. Basically cut the old PV definitions and indent them as a list. You know, like how the miq-template.yaml and miq-template-ext-db.yaml files are now.

• I added a parameters key after the list of objects in the objects top-level-key. I replaced variable-value field values with variables.

I know that might sound confusing. So for example:

    capacity:                                                                                                                                                                                                        
      storage: 15Gi                                                                                                                                                                                       

Becomes

    capacity:
      storage: "${DB_PV_SIZE}"
# ...
parameters:
- name: DB_PV_SIZE
  displayName: PV Size for App DB
  required: true
  description: The size of the MIQ DB PV given in Gi
  value: 15Gi

How to process these? During experimentation/study I figured out that this works:

1. Copy (scp?) your template to your OC cluster
2. oc create -f /path/to/template.yaml - now it's in the system. You can reference it by name when using oc process later on
3. oc process <name-of-template> -p NFS_HOST=nfs.example.com DB_PV_SIZE=15Gi BASE_PATH=/exports | oc create -f - You created the thingies the the thing!

[tbielawa]

@bdunne @fbladilo @carbonin Any more thoughts on this?

[carbonin]

I think @bdunne is traveling this week so I would give him some time to respond.

@fbladilo has no excuse 😜

[tbielawa]

PR sent w/ questions for discussion.

[fbladilo]

@tbielawa @carbonin 😄 Sorry been busy with some ansible training thingies... will check on PR.