[PATCH] Support for Fedora 11 imports

[James Laska]

---
 cobbler/codes.py |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/cobbler/codes.py b/cobbler/codes.py
index 968e517..6da28eb 100644
--- a/cobbler/codes.py
+++ b/cobbler/codes.py
@@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
 ]

VALID_OS_VERSIONS = {
-    "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5", "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24", "generic26", "other" ],
+    "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5", "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11", "generic24", "generic26", "other" ],
     "suse"    : [ "sles10", "generic24", "generic26", "other" ],
     "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
     "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger", "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon", "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],

[Michael DeHaan]

James Laska wrote:
> cobbler/codes.py |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/cobbler/codes.py b/cobbler/codes.py
> index 968e517..6da28eb 100644
> --- a/cobbler/codes.py
> +++ b/cobbler/codes.py
> @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
>  ]
>
> VALID_OS_VERSIONS = {
> "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5", "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24", "generic26", "other" ],
> "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5", "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11", "generic24", "generic26", "other" ],
> "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger", "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon", "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],

Applied to both branches (master for 1.6.5 and devel), thanks!

--Michael

[James Cammarata]

On Tue, 26 May 2009 12:16:08 -0400, Michael DeHaan 
wrote:
> James Laska wrote:
> > cobbler/codes.py |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/cobbler/codes.py b/cobbler/codes.py
> > index 968e517..6da28eb 100644
> > --- a/cobbler/codes.py
> > +++ b/cobbler/codes.py
> > @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
> >  ]
> >
> > VALID_OS_VERSIONS = {
> > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> >
> > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24",
> > "generic26", "other" ],
> > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> >
> > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11",
> > "generic24", "generic26", "other" ],
> >      "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> >      "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> >      "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger",
> >      "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon",
> >      "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],
>
> Applied to both branches (master for 1.6.5 and devel), thanks!

I actually did this over the weekend, and ran into a weird issue.  The
import grabbed the vmlinuz-PAE kernel, but assigned the initrd to be the
standard, non-PAE one.  This caused some weird errors during the kickstart,
basically no drivers were found since it couldn't find the correct
modules.dep file (and no modules directory for the running kernel).  The
vmlinuz/vmlinuz-PAE are in the same directory, causing the issue.

A quick "distro edit" fixed it, however I was wondering if you had tested
using the F11 distro to kickstart something, and if not we may need to make
the import code a little smarter about which initrd/vmlinuz pairs it uses.

[James Laska]

On Tue, 2009-05-26 at 11:22 -0500, James Cammarata wrote:
> On Tue, 26 May 2009 12:16:08 -0400, Michael DeHaan 
> wrote:
> > James Laska wrote:
> > > cobbler/codes.py |    2 +-
> > >  1 files changed, 1 insertions(+), 1 deletions(-)
> > >
> > > diff --git a/cobbler/codes.py b/cobbler/codes.py
> > > index 968e517..6da28eb 100644
> > > --- a/cobbler/codes.py
> > > +++ b/cobbler/codes.py
> > > @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
> > >  ]
> > >
> > > VALID_OS_VERSIONS = {
> > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > >
> > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24",
> > > "generic26", "other" ],
> > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > >
> > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11",
> > > "generic24", "generic26", "other" ],
> > >      "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> > >      "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> > >      "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger",
> > >      "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon",
> > >      "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],
> >
> > Applied to both branches (master for 1.6.5 and devel), thanks!
>
> I actually did this over the weekend, and ran into a weird issue.  The
> import grabbed the vmlinuz-PAE kernel, but assigned the initrd to be the
> standard, non-PAE one.  This caused some weird errors during the kickstart,
> basically no drivers were found since it couldn't find the correct
> modules.dep file (and no modules directory for the running kernel).  The
> vmlinuz/vmlinuz-PAE are in the same directory, causing the issue.
>
> A quick "distro edit" fixed it, however I was wondering if you had tested
> using the F11 distro to kickstart something, and if not we may need to make
> the import code a little smarter about which initrd/vmlinuz pairs it uses.

Thanks James ... I have hit this same issue and worked around it the
same way you've suggested.  I have not had time yet to discern the root
cause as my config.d directory has been inherited from several different
versions of cobbler.

I intend to look eventually, but I suspect we might want to import both
(PAE and non-PAE)?

# cobbler import --name rawhide-20090526 --arch i386
--path /path/to/tree --available-as
http://download.fedora.redhat.com/pub/fedora/linux/development/i386/os

I think users might want to choose the PAE (includes xen) kernel for
certain situations?  So the distros we should create when importing
Fedora 11 trees should be:

- creating new distro: rawhide-20090526-i386
- creating new profile: rawhide-20090526-i386
- creating new profile: rescue-rawhide-20090526-i386

- creating new distro: rawhide-20090526-PAE-i386
- creating new profile: rawhide-20090526-PAE-i386

Does this sound reasonable?

Thanks,
James

[Michael DeHaan]

James Cammarata wrote:
> On Tue, 26 May 2009 12:16:08 -0400, Michael DeHaan 
> wrote:
> > James Laska wrote:
> > > cobbler/codes.py |    2 +-
> > >  1 files changed, 1 insertions(+), 1 deletions(-)
> > >
> > > diff --git a/cobbler/codes.py b/cobbler/codes.py
> > > index 968e517..6da28eb 100644
> > > --- a/cobbler/codes.py
> > > +++ b/cobbler/codes.py
> > > @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
> > >  ]
> > >
> > > VALID_OS_VERSIONS = {
> > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > >
> > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24",
> > > "generic26", "other" ],
> > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > >
> > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11",
> > > "generic24", "generic26", "other" ],
> > >      "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> > >      "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> > >      "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger",
> > >      "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon",
> > >      "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],
> >
> > Applied to both branches (master for 1.6.5 and devel), thanks!
>
> I actually did this over the weekend, and ran into a weird issue.  The
> import grabbed the vmlinuz-PAE kernel, but assigned the initrd to be the
> standard, non-PAE one.  This caused some weird errors during the kickstart,
> basically no drivers were found since it couldn't find the correct
> modules.dep file (and no modules directory for the running kernel).  The
> vmlinuz/vmlinuz-PAE are in the same directory, causing the issue.
>
> A quick "distro edit" fixed it, however I was wondering if you had tested
> using the F11 distro to kickstart something, and if not we may need to make
> the import code a little smarter about which initrd/vmlinuz pairs it uses.

We've had reports of this happening on earlier versions, and I had never 
actually seen it -- couldn't reproduce this at all -- until just very 
recently.

We really need to tweak import to make seperate PAE and non-PAE 
versions, I guess, to do this properly, as much as I hate to fill up the 
distro list.
Thoughts?

--Michael

[James Laska]

On Tue, 2009-05-26 at 14:38 -0400, Michael DeHaan wrote:
> James Cammarata wrote:
> > On Tue, 26 May 2009 12:16:08 -0400, Michael DeHaan 
> > wrote:
> > > James Laska wrote:
> > > > cobbler/codes.py |    2 +-
> > > >  1 files changed, 1 insertions(+), 1 deletions(-)
> > > >
> > > > diff --git a/cobbler/codes.py b/cobbler/codes.py
> > > > index 968e517..6da28eb 100644
> > > > --- a/cobbler/codes.py
> > > > +++ b/cobbler/codes.py
> > > > @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
> > > >  ]
> > > >
> > > > VALID_OS_VERSIONS = {
> > > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > > >
> > > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24",
> > > > "generic26", "other" ],
> > > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > > >
> > > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11",
> > > > "generic24", "generic26", "other" ],
> > > >      "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> > > >      "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> > > >      "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger",
> > > >      "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon",
> > > >      "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],
> > >
> > > Applied to both branches (master for 1.6.5 and devel), thanks!
> >
> > I actually did this over the weekend, and ran into a weird issue.  The
> > import grabbed the vmlinuz-PAE kernel, but assigned the initrd to be the
> > standard, non-PAE one.  This caused some weird errors during the kickstart,
> > basically no drivers were found since it couldn't find the correct
> > modules.dep file (and no modules directory for the running kernel).  The
> > vmlinuz/vmlinuz-PAE are in the same directory, causing the issue.
> >
> > A quick "distro edit" fixed it, however I was wondering if you had tested
> > using the F11 distro to kickstart something, and if not we may need to make
> > the import code a little smarter about which initrd/vmlinuz pairs it uses.
>
> We've had reports of this happening on earlier versions, and I had never 
> actually seen it -- couldn't reproduce this at all -- until just very 
> recently.
>
> We really need to tweak import to make seperate PAE and non-PAE 
> versions, I guess, to do this properly, as much as I hate to fill up the 
> distro list.
> Thoughts?

I don't use them (or don't realize that I do) ... but I agree, I believe
the PAE kernel also provides xen guest support as well.  I believe we
may need to have both distros available after import.

Thanks,
James
> cobbler-devel mailing list
> [e-mail]
> https://fedorahosted.org/mailman/listinfo/cobbler-devel

[Michael DeHaan]

James Laska wrote:
> On Tue, 2009-05-26 at 14:38 -0400, Michael DeHaan wrote:
> > James Cammarata wrote:
> > > On Tue, 26 May 2009 12:16:08 -0400, Michael DeHaan 
> > > wrote:
> > > > James Laska wrote:
> > > > > cobbler/codes.py |    2 +-
> > > > >  1 files changed, 1 insertions(+), 1 deletions(-)
> > > > >
> > > > > diff --git a/cobbler/codes.py b/cobbler/codes.py
> > > > > index 968e517..6da28eb 100644
> > > > > --- a/cobbler/codes.py
> > > > > +++ b/cobbler/codes.py
> > > > > @@ -41,7 +41,7 @@ VALID_OS_BREEDS = [
> > > > >  ]
> > > > >
> > > > > VALID_OS_VERSIONS = {
> > > > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > > > >
> > > > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "generic24",
> > > > > "generic26", "other" ],
> > > > > "redhat"  : [ "rhel2.1", "rhel3", "rhel4", "rhel5", "fedora5",
> > > > >
> > > > > "fedora6", "fedora7", "fedora8", "fedora9", "fedora10", "fedora11",
> > > > > "generic24", "generic26", "other" ],
> > > > >      "suse"    : [ "sles10", "generic24", "generic26", "other" ],
> > > > >      "debian"  : [ "etch", "lenny", "generic24", "generic26", "other" ],
> > > > >      "ubuntu"  : [ "WartyWarthog", "HoaryHedgehog", "BreezyBadger",
> > > > >      "DapperDrake", "EdgyEft", "FeistyFawn", "GutsyGibbon",
> > > > >      "HardyHeron", "IntrepidIbex", "JauntyJackalope" ],
> > > >
> > > > Applied to both branches (master for 1.6.5 and devel), thanks!
> > >
> > > I actually did this over the weekend, and ran into a weird issue.  The
> > > import grabbed the vmlinuz-PAE kernel, but assigned the initrd to be the
> > > standard, non-PAE one.  This caused some weird errors during the kickstart,
> > > basically no drivers were found since it couldn't find the correct
> > > modules.dep file (and no modules directory for the running kernel).  The
> > > vmlinuz/vmlinuz-PAE are in the same directory, causing the issue.
> > >
> > > A quick "distro edit" fixed it, however I was wondering if you had tested
> > > using the F11 distro to kickstart something, and if not we may need to make
> > > the import code a little smarter about which initrd/vmlinuz pairs it uses.
> >
> > We've had reports of this happening on earlier versions, and I had never 
> > actually seen it -- couldn't reproduce this at all -- until just very 
> > recently.
> >
> > We really need to tweak import to make seperate PAE and non-PAE 
> > versions, I guess, to do this properly, as much as I hate to fill up the 
> > distro list.
> > Thoughts?
>
> I don't use them (or don't realize that I do) ... but I agree, I believe
> the PAE kernel also provides xen guest support as well.  I believe we
> may need to have both distros available after import.
>
> Thanks,
> James
> > cobbler-devel mailing list
> > [e-mail]
> > https://fedorahosted.org/mailman/listinfo/cobbler-devel
> >
> > cobbler-devel mailing list
> > [e-mail]
> > https://fedorahosted.org/mailman/listinfo/cobbler-devel

This is now implemented on both branches.

The master branch (1.6.5) will also be able to do F-11 imports and get 
this correct.

--Michael

[James Cammarata]

On Tue, 26 May 2009 18:03:05 -0400, Michael DeHaan 
wrote:
> This is now implemented on both branches.
>
> The master branch (1.6.5) will also be able to do F-11 imports and get 
> this correct.
>
> --Michael

That was fast :) I was going to look into this later tonight.

It seemed pretty simple to me, basically it was looking for vmlinuz*, so it
was kind of a race condition with globbing as to which would be the one
that was selected when the first initrd was found.  For those who haven't
looked at the distro import code, essentially it does a recursive descent
on the distro directory.  It keeps track of any initrd/vmlinuz files it has
found, and as soon as it has found one of each it says DONE! and adds the
distro.  To fix, just stop letting it look for vmlinuz*.

[Michael DeHaan]

James Cammarata wrote:
> On Tue, 26 May 2009 18:03:05 -0400, Michael DeHaan 
> wrote:
> > This is now implemented on both branches.
> >
> > The master branch (1.6.5) will also be able to do F-11 imports and get 
> > this correct.
> >
> > --Michael
>
> That was fast :) I was going to look into this later tonight.
>
> It seemed pretty simple to me, basically it was looking for vmlinuz*, so it
> was kind of a race condition with globbing as to which would be the one
> that was selected when the first initrd was found.  For those who haven't
> looked at the distro import code, essentially it does a recursive descent
> on the distro directory.  It keeps track of any initrd/vmlinuz files it has
> found, and as soon as it has found one of each it says DONE! and adds the
> distro.  To fix, just stop letting it look for vmlinuz*.

Yeah, pretty much. So right now instead of waiting for "one of each" to 
be filled up, we keep track of two
seperate pairs, and wait for either pair of vmlinuz/initrds to fill up.

Right now we create both the PAE and non-PAE version as both distro and 
profile objects, if found.