On 07.08.2015 16:16, Daniel Henrique Barboza wrote:


On 08/07/2015 11:12 AM, Daniel Henrique Barboza wrote:


On 08/07/2015 05:53 AM, Walter Niklaus wrote:
Aline,

the goal of my mail wasn't a scrum summary only, I was rather trying to get us into a design thinking mode:  looking at the problem from a user point of view, what would a user expect to exploit in the various use cases.  I'll keep focusing on this aspect in our future discussions :-)

I like your proposal from an overall point because for making progress now it is the best solution.
When it comes to plugins-priority and implentation I have a simpler proposal:
  - Instead of this smart logic you are proposing implement the following:
        - whenever common-host plugin is available it will get installed and create a Host tab with the functionality 2,3 and 6
        - when Ginger plugin is getting installed it will simply extend the Host tab with all the other functionalties

Why get rid of the 'Administration' tab from Ginger? We can use the 'Host' tab to provide basic
information + debug reports and the Administration tab to manage the host itself. We can even
rename the 'Host' tab to 'Host status' to clarify even further.


Administration isn't really specific about being host management. I guess the only aspect matching this description may be User Management and all these aspects we are discussing are Host related.

Get rid of the Ginger tab can, and will, confuse the existing users. They will install the plug-in,
they will not see the Administration tab and will think that Ginger installation failed.

Can we document this ? And we are talking about a new release, based on a new UI-Design.
Otherwise we would have to stick to current UI design and layout forever.  Applying design thinking may introduce some additional changes but I feel as long as we are doing a good job in providing an intuitive UI, the user will adapt very easy.


Besides that, the current engine does not support a plug-in that extends an existing tab from
another plug-in. I have no idea of the amount of work this can take.

I would hope that with the new UI design, this restriction isn't there anymore. Of course we have to look at the effort.
But I would like to exploit this feature in other situations as well since I'm planning to have the IO-management for System z as a separate plugin.




On 06.08.2015 17:11, Aline Manera wrote:

Thanks, Walter, to send the scrum meeting summary here!

Here are my thoughts on all that.

First, let me clarify the proposal of each piece of cake! =)

A) Wok is a *generic web server framework based on plugins*.
    By generic, I mean it should only expose APIs and functionalities required for a web server. Login, logout, plugins support, i18n support, message error handling and much more.

B) Kimchi is a wok plugin for virtual machine management.
    And it is independent system platform: x86, Power or Z.

C) Ginger is a wok plugin for host management.
    And it is independent system platform: x86, Power or Z.

By now, it is all we have. So I'd like to concentrate our effort on it.

Now thinking about which features from Kimchi Host tab can be moved to Ginger.
Let do it item by item. The Host Tab is composed by:

1) Restart, Shutdown, Connect operations
    I don't see those functionalities close related to virtual machines management. So for me, it is fine and good to move them to Ginger.


2) Basic Information
    The kind of information may be very useful to user while manage virtual machine. Specially by the amount of memory and number of CPUs.
    With that information the user can properly balance his/her virtual machines configuration to have better system performance.

3) Host statistics
    The same I described on item 2.

4) Software Update
    I don't see this functionality close related to virtual machines management. So for me, it is fine and good to move them to Ginger.

5) Repository management
    I don't see this functionality close related to virtual machines management. So for me, it is fine and good to move them to Ginger.

6) Debug reports
    This functionality may be interesting for virtual machine management when some of them represents a problem or bad performance.
    So user can easily grab the system logs to check what is going wrong.

So if no one opposes, we can start moving 1, 4 and 5 to Ginger.
As we have 2 different open source communities we need to coordinate that work. Initially the patch for Kimchi will be for removing those features and to Ginger to add them.
The Kimchi patch must be simpler but it will require more work on Ginger side.

About 2, 3 and 6: I really understand how those information are important for virtual machine management and also for host management, i.e, Kimchi and Ginger.
(And I hope you all do the same :-) )

So in my mind, we are discussing a solution to expose those information on both, Kimchi and Ginger, without making it duplicated somehow to user.

As per discussion (see item A), wok is a generic framework and should not handle those kind of APIs. (Agree?)
And also it should not have any default plugin (otherwise, we could continue having Kimchi as default without the need to have the wok framework)
While loading wok without any plugin, it should display a simple page "Welcome to wok" or something like that but without any functionality.

So my proposal is to create a new and simple plugin (let's call it as common-host plugin) that expose those APIs without any UI.

Kimchi and Ginger will have a dependency on this common-host plugin and will provide the proper UI for it.

To do not duplicate information while loading Kimchi and Ginger together, I propose to add a smart logic for it:

- Kimchi will always load the Host tab with the common-host plugin (as it is today).
- Ginger will load the common-host plugin *only and if only* it is running standalone.

What do you think about it?

Regards,
Aline Manera

On 06/08/2015 11:15, Walter Niklaus wrote:
Picking up the discussion from the Scrum meeting about where (which plugin) certain functionalities should be.

To make sure we don't miss this aspect, I'm re-iterating on the high level use cases.
Currently I see the following major usecases now and in the near future (next year):
  1.  A user wants to perform base Linux management only.
        - here he needs all the generic Host-Management functionality + the platform specific stuff like:
          Power-FW code update, Energy Management on Power or IO-Device Management on System z
  2.  A user wants to manage KVM Virtual Machines.
       - his primary scope are VMs. How much of the Host and Platform specific Management functionality is required here ?
  3.  A user wants to manage Containers.
       - his primary scope are Container. How much of the Host and Platform specific Management functionality is required here ?
  4.  A user wants to manage Containers and KVM Virtual Machines .
      - his primary scope are Container and VMs. How much of the Host and Platform specific Management functionality is required here ?

Our current discussion now is for the usecases 2,3 and 4: How much of the Host and Platform specific Management functionality is required and what's the best way to organize and package it.
One possibility could be to have all Host-Management functionality looked at being part of the default/basic functionset and delivery and have the Platform specifc functionality as optional plugins. The disadvantage of this approach would be that all the following functionality:
Basic Information, System Statistics, Network  (Host NICs,DNS ...), Storage/SAN (Host Storage), User Management, Configuration Backup, Software Updates, Repositories, Debug Reports would be present in the Container and VM usecases by default.
Do we know what a user really needs and wants in the usecases 2,3 and 4 ?  I guess this depends to a large degree of the toolset she/he is using beside Kimchi and Ginger. If there is no other tooling available she/he may be happy about the shipped functionset, but for sure there are other situation where she/he may not be interested in some of the functionality.

What could be the reasons a user would want to pick selectively ?
  a.   functionality not required or maybe even conflicting with some other tooling: for example Software Updates
        are managed from some central instance
  b.   installing a reduced functionset could reduce the external package dependencies and could reduce the amount of updates
  c.   simplification on the UI by eliminating unrequired stuff

Ideally the user could choose on an individual functionality base and configure the tool based on his needs.
I guess satisfying the reasons a. and c. from above could be implemented via UI customisation even on an individual Kimchi/Ginger user base.
Reason b. can be probably achieved only by segregating the set of fuctionality in separate plugins.



On 04.08.2015 17:26, Walter Niklaus wrote:
... Daniel sorry for the duplicate send, I missed to reply to all so the mail didn't go to the mailing list.
 
On 04.08.2015 14:39, Daniel Henrique Barboza wrote:


On 08/04/2015 04:56 AM, Walter Niklaus wrote:
Hi Daniel,

sorry for missing the thread where this topic was discussed.

I can fully understand the point about Basic Information and System Statistics being relevant for Virtualization management as well and I like the idea of potentially making it part of the base framework because they would be very usefull for other plugins, like Container-Management as well.
The interesting question is then if some of the other functions wouldn't make sense to be part of the basic framework as well. Debug reports would be a classical candidate from my point of view, but wouldn't some of the other functions be usefull in the base as well ?

If we're really going in that approach (putting basic features in WoK), I agree. We would have to
discuss each existing feature and evaluate if it belongs to kimchi, ginger or wok.

I guess we really need to have a discussion on the individual features but I would like to start this one from a user requirements point of view.
Currently I see the following major usecases now and in the near feature:
  1.  A user wants to perform base Linux management only.
  2.  A user wants to manage KVM Virtual Machines.
  3.  A user wants to manage Containers.
  4.  A user wants to manage Containers and KVM Virtual Machines .

For the usecases 2, 3 and 4 the user needs usecase 1 as well in order to prepare and manage the Host machine.

I'm not proposing to make the Linux Host Management part of the base framework because we just separated out Kimchi of it, but I think it makes a lot of sense to deliver the Host Management plugin by default with the base framework.



Looking at the problem form a different angle: wouldn't it make sense to package and deliver the base framework with the Ginger plugin by default because the Host-functionality Ginger is offering would be usefull for the other plugins like Virtualization and Containers ?

What I missed in my previous mail is the aspect about platform specific functionality. This functionality, like PPC firmware update or IO-device management for Linux on z should be made available as individual plugins.

At this moment Ginger can handle multi-arch features fairly well. For example, Firmware
Update does not appear when running the plug-in in an Intel computer. The feature you mentioned,
IO-device management for Linux on Z, would be available only when running Ginger in a Linux
for Z host.

There's absolutely nothing holding you from making a brand new plug-in for the Z features instead
of adding them to Ginger, but it is important to know that Ginger is designed for these scenarios.
You can even create a new UI tab in Ginger, something like 'Z management' which would contain all Z related features. This tab would only appear in a Linux on Z host. From the UI perspective it looks
like a brand new plug-in working together with Ginger common features in the 'Administration' tab.


Please let me know what you think about this option.

Thanks,
Walter.


On 03.08.2015 18:51, Daniel Henrique Barboza wrote:
Hi Walter,

We've had this discussion with the community a few months ago in the thread

"[RFC] Moving some features of Host tab to Ginger"

And we agreed to start it by moving only Software Update, Repositories and
Debug Reports from Kimchi to Ginger.

The Basic Information and System Statistics can't be taken away from Kimchi because there
are relevant information for the creation of VMs there, such as Memory Available. But I agree
that these information fits nicely in Ginger too.

One alternative (just came in my head now) is to move these "neutral" functions
to a "Basic System Info" in WoK. That way both Kimchi and Ginger users can access
the information.


Thanks,


Daniel

On 08/03/2015 12:15 PM, Walter Niklaus wrote:

After separating out Kimchi as an indvidual plugin from the base
framework it would be great to have a clean separation between Host- and
Virtualization Management functions. I'm planning to work on this topic
in the next few weeks and have prepared a proposal of the functionsplit.
Plugin functionality:
     -  Ginger:
           - Basic Information
           - System Statistics
           - Network  (Host NICs)
           - Storage/SAN (Host Storage)
           - User Management
           - Configuration Backup
           - Software Updates
           - Repositories
           - Debug Reports
           - PPC related functions:  Firmware Update & Power Management
     - Kimchi:
           -  Templates
           -  Guests
           -  Networks (virtual)
           -  Storage (Pools for VMs)

Since there are plans to restructure the UI for one of the next
releases, I'm proposing to do only some minimal investments in
reflecting this new finctionsplit. Therefore I'm proposing to make the
Host tab as the one and only Tab for Ginger and move everything from the
Administration Tab into the Host Tab.  This would be just an
intermediate solution till we implement the new UI design. Please see
the attached PDF.
Thanks in advance for your feedback.

Walter.











_______________________________________________
Kimchi-devel mailing list
Kimchi-devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/kimchi-devel



_______________________________________________
Kimchi-devel mailing list
Kimchi-devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/kimchi-devel






_______________________________________________
Kimchi-devel mailing list
Kimchi-devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/kimchi-devel




_______________________________________________
Kimchi-devel mailing list
Kimchi-devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/kimchi-devel