[node-devel] Storage Stuff, Installation Module and Anaconda

[Fabian Deutsch]

Hey,

this has been discussed a couple of times over the last months (already!
): To use some parts of anaconda (python-blivet) in our storage and
installer code-base.

While investigating this a bit further on last Friday I started to
wonder if we can't even use more parts of anaconda, basically ending in
a thought where I wondered how much it would take to make anaconda
capable of doing the Node installation.


The biggest issue with anaconda right now is that the "modus operandi"
is around files. Our method on the other hand is around installing
images.

If we can teach anaconda to install images then I could imagine the
following workflow:

1. Our installer create a ks
2. anaconda does the installation based on ks

So I would not actually expose anacondas UI.
But - We could possibly also leverage other parts of anaconda, e.g.
basic network configuration stuff, and maybe even more.

I have not yet investigate how much more dependencies we'd be pulling
in.

It's not that I want to go down this way right now. It just came to my
selection of possibilities.

Greetings so far
- fabian

[Jiri Moskovcak]

Post by Fabian Deutsch
Hey,
this has been discussed a couple of times over the last months (already!
): To use some parts of anaconda (python-blivet) in our storage and
installer code-base.
While investigating this a bit further on last Friday I started to
wonder if we can't even use more parts of anaconda, basically ending in
a thought where I wondered how much it would take to make anaconda
capable of doing the Node installation.
The biggest issue with anaconda right now is that the "modus operandi"
is around files. Our method on the other hand is around installing
images.
If we can teach anaconda to install images then I could imagine the
1. Our installer create a ks
2. anaconda does the installation based on ks
So I would not actually expose anacondas UI.
But - We could possibly also leverage other parts of anaconda, e.g.
basic network configuration stuff, and maybe even more.
I have not yet investigate how much more dependencies we'd be pulling
in.
It's not that I want to go down this way right now. It just came to my
selection of possibilities.
Greetings so far
- fabian

Hi Fabian,
it's quite an interesting idea and it actually might be easier than you
think. You should be able to create an rpm package from the image and
just create anaconda install disc with that rpm on it and let anaconda
do it's job without having to change it's backend to understand images.

--Jirka

Post by Fabian Deutsch
_______________________________________________
node-devel mailing list
http://lists.ovirt.org/mailman/listinfo/node-devel

[Jiri Moskovcak]

Post by Jiri Moskovcak

Post by Fabian Deutsch
Hey,
this has been discussed a couple of times over the last months (already!
): To use some parts of anaconda (python-blivet) in our storage and
installer code-base.
While investigating this a bit further on last Friday I started to
wonder if we can't even use more parts of anaconda, basically ending in
a thought where I wondered how much it would take to make anaconda
capable of doing the Node installation.
The biggest issue with anaconda right now is that the "modus operandi"
is around files. Our method on the other hand is around installing
images.
If we can teach anaconda to install images then I could imagine the
1. Our installer create a ks
2. anaconda does the installation based on ks
So I would not actually expose anacondas UI.
But - We could possibly also leverage other parts of anaconda, e.g.
basic network configuration stuff, and maybe even more.
I have not yet investigate how much more dependencies we'd be pulling
in.
It's not that I want to go down this way right now. It just came to my
selection of possibilities.
Greetings so far
- fabian

Hi Fabian,
it's quite an interesting idea and it actually might be easier than you
think. You should be able to create an rpm package from the image and
just create anaconda install disc with that rpm on it and let anaconda
do it's job without having to change it's backend to understand images.
--Jirka

- and actually when you're installing from livecd you're installing an
image, so the logic is already there..

--Jirka

Post by Jiri Moskovcak

Post by Fabian Deutsch
_______________________________________________
node-devel mailing list
http://lists.ovirt.org/mailman/listinfo/node-devel

_______________________________________________
node-devel mailing list
http://lists.ovirt.org/mailman/listinfo/node-devel

[Fabian Deutsch]

Post by Jiri Moskovcak

Post by Jiri Moskovcak

Post by Fabian Deutsch
Hey,
this has been discussed a couple of times over the last months (already!
): To use some parts of anaconda (python-blivet) in our storage and
installer code-base.
While investigating this a bit further on last Friday I started to
wonder if we can't even use more parts of anaconda, basically ending in
a thought where I wondered how much it would take to make anaconda
capable of doing the Node installation.
The biggest issue with anaconda right now is that the "modus operandi"
is around files. Our method on the other hand is around installing
images.
If we can teach anaconda to install images then I could imagine the
1. Our installer create a ks
2. anaconda does the installation based on ks
So I would not actually expose anacondas UI.
But - We could possibly also leverage other parts of anaconda, e.g.
basic network configuration stuff, and maybe even more.
I have not yet investigate how much more dependencies we'd be pulling
in.
It's not that I want to go down this way right now. It just came to my
selection of possibilities.
Greetings so far
- fabian

Hi Fabian,
it's quite an interesting idea and it actually might be easier than you
think. You should be able to create an rpm package from the image and
just create anaconda install disc with that rpm on it and let anaconda
do it's job without having to change it's backend to understand images.
--Jirka

- and actually when you're installing from livecd you're installing an
image, so the logic is already there..

Hey Jirka,

thanks for your input!

As far as I understand the LiveCd Payload it looks like the filesystem
tree is copied, and not the image itself.

Greetings
fabian

[Fabian Deutsch]

Post by Jiri Moskovcak

Post by Fabian Deutsch
Hey,
this has been discussed a couple of times over the last months (already!
): To use some parts of anaconda (python-blivet) in our storage and
installer code-base.
While investigating this a bit further on last Friday I started to
wonder if we can't even use more parts of anaconda, basically ending in
a thought where I wondered how much it would take to make anaconda
capable of doing the Node installation.
The biggest issue with anaconda right now is that the "modus operandi"
is around files. Our method on the other hand is around installing
images.
If we can teach anaconda to install images then I could imagine the
1. Our installer create a ks
2. anaconda does the installation based on ks
So I would not actually expose anacondas UI.
But - We could possibly also leverage other parts of anaconda, e.g.
basic network configuration stuff, and maybe even more.
I have not yet investigate how much more dependencies we'd be pulling
in.
It's not that I want to go down this way right now. It just came to my
selection of possibilities.
Greetings so far
- fabian

Hi Fabian,
it's quite an interesting idea and it actually might be easier than you
think. You should be able to create an rpm package from the image and
just create anaconda install disc with that rpm on it and let anaconda
do it's job without having to change it's backend to understand images.

Hey,

IIUYC I am not sure that this would solve Node's installation problem.

Because Node is an image (ISO) which is copied onto a Logical Volume
(LV). A seperate boot partition then points to the LV at boot time.

And this part - writing an image to a LV during installation - is what
might not be possible right now.

But maybe it can be added.

Greetings
fabian