On 2/22/2014 21:37, Chris Murphy wrote:
On Feb 22, 2014, at 12:46 PM, Don Levey <fedora-list(a)the-leveys.us>
wrote:
> On 2/21/2014 11:38, Chris Murphy wrote:
...
>
>> What do you get for:
>>
>> mdadm -E /dev/sda mdadm -E /dev/sdc
>>
>>
> (this time, with commands, because I can learn. :-) )
>
> [root@localhost don]# mdadm -E /dev/sda /dev/sda: MBR Magic : aa55
> Partition[0] : 102760449 sectors at 63 (type 07)
> Partition[1] : 409600 sectors at 102760512 (type 83)
> Partition[2] : 209326268 sectors at 103170112 (type 8e)
> [root@localhost don]# mdadm -E /dev/sdc /dev/sdc: MBR Magic : aa55
> Partition[0] : 102760449 sectors at 63 (type 07)
> Partition[1] : 409600 sectors at 102760512 (type 83)
> Partition[2] : 209326268 sectors at 103170112 (type 8e)
>
> So yes, they seem to be relating the same thing.
It's not finding the IMSM metadata if this is firmware raid. In the
BIOS setup, you should find some setting related to this. Typically
the option setting is either raid, ide, or ahci. Maybe it was set to
raid when you initially did the install and then got changed to ide?
I wouldn't change it from what it is now, but it's worth checking the
setting to see if you recall it being changed recently.
I will check when I can reboot next - hopefully tonight.
Also show the result from each:
cat /proc/mdstat pvck /dev/sda3 pvck /dev/sdc3
[root@localhost don]# cat /proc/mdstat
Personalities :
unused devices: <none>
[root@localhost don]# pvck /dev/sda3
WARNING: Failed to connect to lvmetad: No such file or directory.
Falling back to internal scanning.
Found label on /dev/sda3, sector 1, type=LVM2 001
Found text metadata area: offset=4096, size=192512
[root@localhost don]# pvck /dev/sdc3
WARNING: Failed to connect to lvmetad: No such file or directory.
Falling back to internal scanning.
Found label on /dev/sdc3, sector 1, type=LVM2 001
Found text metadata area: offset=4096, size=192512
What I'm betting on at this point (guessing), is that the raid
array
is not active, and therefore two identical UUID physical volumes are
present rather than appearing as one, and grub is becoming confused.
On a separate machine, running software RAID1, I see active raid
devices. That I am not seeing them here would support your suggestion
that the raid array isn't active, no?
And the other problem is that there simply isn't enough space to
embed core.img in the MBR gap as usual, because the support modules
for RAID and LVM make it too big. There is a way to use block lists,
but before doing that the whole raid thing needs to be figured out.
Interesting. This makes me wonder why it all worked before.
This was a machine I inherited from a friend, with the raid devices
already there, and running Windows. When I installed on it, from what I
recall, it presented me with a single volume, and never behaved as if
there were two devices.
I haven't done a lot of testing in this area with actual hardware
but
some people say using software raid between Windows and Linux is
flakey. I think it ought to work or it's a bug. But the thing to
realize is that you have a Windows driver doing software raid when
Windows is running; and then you have a linux driver doing software
raid when linux is running. So it's absolutely got to work correctly
or you've totally defeated the purpose of even having raid1, and
instead you're better off with just a regular consistent backup, like
once an hour or something.
I can appreciate that. If this were a mission-critical machine I might
be able to justify the setup for a regular, hourly, backup. At this
point, though, that seems overkill. If this hadn't been working before,
I would probably just accept that it's not going to work and move on.
Thank you again,
-Don