Write an AUTO ... line to mdadm.conf (#537329)
 

Write an "AUTO +imsm +1.x -all" line to mdadm.conf, and always
write mdadm.conf. This is necessary for proper activation of
Intel BIOS RAID sets when they are configured in the OROM
after installation.

This patch is intended for f13-branch and rhel6-branch too.
---
storage/__init__.py | 11 +++--------
1 files changed, 3 insertions(+), 8 deletions(-)

diff --git a/storage/__init__.py b/storage/__init__.py
index ebbf9ed..b19dd18 100644
--- a/storage/__init__.py
+++ b/storage/__init__.py
@@ -2135,7 +2135,6 @@ class FSSet(object):

def mdadmConf(self):
""" Return the contents of mdadm.conf. """
- retval = None
arrays = self.devicetree.getDevicesByType("mdarray")
arrays.extend(self.devicetree.getDevicesByType("md biosraidarray"))
arrays.extend(self.devicetree.getDevicesByType("md container"))
@@ -2148,19 +2147,15 @@ class FSSet(object):

conf = "# mdadm.conf written out by anaconda
"
conf += "MAILADDR root
"
+ conf += "AUTO +imsm +1.x -all
"
devices = self.mountpoints.values() + self.swapDevices
for array in arrays:
- writeConf = False
for device in devices:
if device == array or device.dependsOn(array):
- writeConf = True
+ conf += array.mdadmConfEntry
break

- if writeConf:
- conf += array.mdadmConfEntry
- retval = conf
-
- return retval
+ return conf

def multipathConf(self):
""" Return the contents of multipath.conf. """
--
1.7.0.1

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Write an AUTO ... line to mdadm.conf (#537329)
 

> Write an "AUTO +imsm +1.x -all" line to mdadm.conf, and always
> write mdadm.conf. This is necessary for proper activation of
> Intel BIOS RAID sets when they are configured in the OROM
> after installation.

ACK to both.

- Chris

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Write an AUTO ... line to mdadm.conf (#537329)
 

On Tue, 2010-04-13 at 17:11 +0200, Hans de Goede wrote:
> Write an "AUTO +imsm +1.x -all" line to mdadm.conf, and always
> write mdadm.conf. This is necessary for proper activation of
> Intel BIOS RAID sets when they are configured in the OROM
> after installation.
>
> This patch is intended for f13-branch and rhel6-branch too.
> ---
> storage/__init__.py | 11 +++--------
> 1 files changed, 3 insertions(+), 8 deletions(-)
>
> diff --git a/storage/__init__.py b/storage/__init__.py
> index ebbf9ed..b19dd18 100644
> --- a/storage/__init__.py
> +++ b/storage/__init__.py
> @@ -2135,7 +2135,6 @@ class FSSet(object):
>
> def mdadmConf(self):
> """ Return the contents of mdadm.conf. """
> - retval = None
> arrays = self.devicetree.getDevicesByType("mdarray")
> arrays.extend(self.devicetree.getDevicesByType("md biosraidarray"))
> arrays.extend(self.devicetree.getDevicesByType("md container"))
> @@ -2148,19 +2147,15 @@ class FSSet(object):
>
> conf = "# mdadm.conf written out by anaconda
"
> conf += "MAILADDR root
"
> + conf += "AUTO +imsm +1.x -all
"

This makes me nervous. What else are we changing here besides firmware
raid? Don't say nothing -- I won't believe you.

Dave

> devices = self.mountpoints.values() + self.swapDevices
> for array in arrays:
> - writeConf = False
> for device in devices:
> if device == array or device.dependsOn(array):
> - writeConf = True
> + conf += array.mdadmConfEntry
> break
>
> - if writeConf:
> - conf += array.mdadmConfEntry
> - retval = conf
> -
> - return retval
> + return conf
>
> def multipathConf(self):
> """ Return the contents of multipath.conf. """


_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Write an AUTO ... line to mdadm.conf (#537329)
 

Hi,

On 04/13/2010 08:54 PM, David Lehman wrote:

On Tue, 2010-04-13 at 17:11 +0200, Hans de Goede wrote:

Write an "AUTO +imsm +1.x -all" line to mdadm.conf, and always
write mdadm.conf. This is necessary for proper activation of
Intel BIOS RAID sets when they are configured in the OROM
after installation.

This patch is intended for f13-branch and rhel6-branch too.
---
storage/__init__.py | 11 +++--------
1 files changed, 3 insertions(+), 8 deletions(-)

diff --git a/storage/__init__.py b/storage/__init__.py
index ebbf9ed..b19dd18 100644
--- a/storage/__init__.py
+++ b/storage/__init__.py
@@ -2135,7 +2135,6 @@ class FSSet(object):

def mdadmConf(self):
""" Return the contents of mdadm.conf. """
- retval = None
arrays = self.devicetree.getDevicesByType("mdarray")
arrays.extend(self.devicetree.getDevicesByType("md biosraidarray"))
arrays.extend(self.devicetree.getDevicesByType("md container"))
@@ -2148,19 +2147,15 @@ class FSSet(object):

conf = "# mdadm.conf written out by anaconda
"
conf += "MAILADDR root
"
+ conf += "AUTO +imsm +1.x -all
"


This makes me nervous. What else are we changing here besides firmware
raid? Don't say nothing -- I won't believe you.



We are telling mdadm that it is ok to auto assembly imsm (Intel firmware RAID)
*and* normal mdraid sets with a metadata version of 1.x, which are sets
either manually created by people, or created by the F-13 (or later) anaconda.

The idea here is that 1.x metadata contains a homehost identifier, so if some
random disk with mdraid 1.x metadata is plugged in, the homehost won't match
and thus it won't be auto assembled, where as sets created on the host which
is booting will be. This has all been discussed in the "mdadm update" mail
thread on which you are on the CC (although I cannot blame you for not following
it).

This is done an the explicit request of Doug Ledford, and I double checked with
him that the +1.x should be added for F-13 too.

Note that this only impacts post installation booting, and not anything happening
while we are installing.

Regards,

Hans

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Write an AUTO ... line to mdadm.conf (#537329)
 

On Tue, 2010-04-13 at 21:57 +0200, Hans de Goede wrote:
> Hi,
>
> On 04/13/2010 08:54 PM, David Lehman wrote:
> > On Tue, 2010-04-13 at 17:11 +0200, Hans de Goede wrote:
> >> Write an "AUTO +imsm +1.x -all" line to mdadm.conf, and always
> >> write mdadm.conf. This is necessary for proper activation of
> >> Intel BIOS RAID sets when they are configured in the OROM
> >> after installation.
> >>
> >> This patch is intended for f13-branch and rhel6-branch too.
> >> ---
> >> storage/__init__.py | 11 +++--------
> >> 1 files changed, 3 insertions(+), 8 deletions(-)
> >>
> >> diff --git a/storage/__init__.py b/storage/__init__.py
> >> index ebbf9ed..b19dd18 100644
> >> --- a/storage/__init__.py
> >> +++ b/storage/__init__.py
> >> @@ -2135,7 +2135,6 @@ class FSSet(object):
> >>
> >> def mdadmConf(self):
> >> """ Return the contents of mdadm.conf. """
> >> - retval = None
> >> arrays = self.devicetree.getDevicesByType("mdarray")
> >> arrays.extend(self.devicetree.getDevicesByType("md biosraidarray"))
> >> arrays.extend(self.devicetree.getDevicesByType("md container"))
> >> @@ -2148,19 +2147,15 @@ class FSSet(object):
> >>
> >> conf = "# mdadm.conf written out by anaconda
"
> >> conf += "MAILADDR root
"
> >> + conf += "AUTO +imsm +1.x -all
"
> >
> > This makes me nervous. What else are we changing here besides firmware
> > raid? Don't say nothing -- I won't believe you.
> >
>
> We are telling mdadm that it is ok to auto assembly imsm (Intel firmware RAID)
> *and* normal mdraid sets with a metadata version of 1.x, which are sets
> either manually created by people, or created by the F-13 (or later) anaconda.
>
> The idea here is that 1.x metadata contains a homehost identifier, so if some
> random disk with mdraid 1.x metadata is plugged in, the homehost won't match
> and thus it won't be auto assembled, where as sets created on the host which

The imsm part is fine with me, but not the 1.x. You could argue that it
depends on the whole-system philosophy on automatic assembly/activation
of high-level devices like lvm and md. I'm fairly certain that lvm does
not do auto-activation. More to the point, this activates only md
devices that can reasonably be believed to be local. What about the
other ones (older metadata, non-matching homehost)? Am I crazy for
caring that we seem to be aiming for as little consistency as is humanly
possible?

>
> Note that this only impacts post installation booting, and not anything happening
> while we are installing.

Right, so go ahead and push it if you want. I don't have to like the way
md behaves after installation.

Dave

>
> Regards,
>
> Hans
>
> _______________________________________________
> Anaconda-devel-list mailing list
> Anaconda-devel-list@redhat.com
> https://www.redhat.com/mailman/listinfo/anaconda-devel-list


_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Write an AUTO ... line to mdadm.conf (#537329)
 

Hi,

On 04/14/2010 01:04 AM, David Lehman wrote:
<snip>


This makes me nervous. What else are we changing here besides firmware
raid? Don't say nothing -- I won't believe you.



We are telling mdadm that it is ok to auto assembly imsm (Intel firmware RAID)
*and* normal mdraid sets with a metadata version of 1.x, which are sets
either manually created by people, or created by the F-13 (or later) anaconda.

The idea here is that 1.x metadata contains a homehost identifier, so if some
random disk with mdraid 1.x metadata is plugged in, the homehost won't match
and thus it won't be auto assembled, where as sets created on the host which


The imsm part is fine with me, but not the 1.x. You could argue that it
depends on the whole-system philosophy on automatic assembly/activation
of high-level devices like lvm and md. I'm fairly certain that lvm does
not do auto-activation.


Goin a bit off topic here, but lvm actually does assemble anything under the
sun which remotely looks like lvm on boot up.


More to the point, this activates only md
devices that can reasonably be believed to be local. What about the
other ones (older metadata, non-matching homehost)? Am I crazy for
caring that we seem to be aiming for as little consistency as is humanly
possible?



I don't feel this is my call to make. If Doug wants this to be the defaults for
mdraid, so be it. I'm just following his wishes here, this policy needs to be in
anaconda as anaconda (over)writes the config file, but it really is mdadm policy
not anaconda policy.

Regards,

Hans

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list