Re: Linux software raid troubles

[Reindl Harald <h.reindl@thelounge.net>]

Am 12.04.2017 um 17:36 schrieb linuxknight:
> Thank you Reindl, Using your method would I be able to apply this IMG
> file to a fresh raid1 mirror and still have it be bootable?

that's the whole point - there is no difference if you have another 
phyiscal disk or a image-file as destination - thanks linux everything 
is a file

whenever you play around with disks which might fail or are already 
broken take a complete image as soon as possible because before you try 
to restore something from that image you can even copy that one, try to 
mount it, play around and whenever you are unsure if you damaged it's 
state just make a fresh copy from the untouched first backup

> The reason I ask is I was looking at this guide,
> https://www.data-medics.com/forum/how-to-clone-a-hard-drive-with-bad-sectors-using-ddrescue-t133.html
> It has a method to transfer drive to drive.  I was thinking I would
> create the fresh RAID mirror on the dedicated LSI card, then ddrescue
> possibly bad drive to the new raid mirror.  Is this a bad idea?
> 
> On Wed, Apr 12, 2017 at 11:29 AM, Reindl Harald <h.reindl@thelounge.net> wrote:
>> please no private-only respones
>>
>> Am 12.04.2017 um 16:52 schrieb linuxknight:
>>>
>>> Thanks you for the reply.  I was just examining the hardward in my
>>> server and it looks like there is an LSI card in there.  If I create a
>>> new Hardware raid mirror in that controller, is it possible to use the
>>> ddrescue to get my current OS onto that mirror and boot from it?  Im
>>> unfamiliar with the ddrescue but will certainly read up more.
>>
>>
>> "ddrescue" is at the end of the day the same as "dd"
>>
>> it reads the whole drive block-by-block and writes it to a image file, later
>> you can do "dd if=image.mig of=/dev/sdX bs=1M" and you get a 100% identical
>> state of the disk
>>
>> so just put out that drive, connect it to a ordinary SATA adapter, take the
>> image and be happy that you have a backup, if the RAID-controller has stored
>> whatever metadata on begin of the drive it's also part of the image
>>
>> and hence leave out that controller to get a 100% block-by-block copy of the
>> whole drive
>>
>>
>>> On Wed, Apr 12, 2017 at 10:45 AM, Reindl Harald <h.reindl@thelounge.net>
>>> wrote:
>>>>
>>>>
>>>>
>>>> Am 12.04.2017 um 16:31 schrieb linuxknight:
>>>>>
>>>>>
>>>>> Last weekend I was moving a server with a raid1 configuration,
>>>>> controlled by a Intel Corporation 82801 SATA RAID Controller.  Upon
>>>>> reboot I noticed the degraded message (server hadnt been rebooted in a
>>>>> couple years).
>>>>>
>>>>> The raid1 array was two 500gb black WD drives.  I wasnt able to locate
>>>>> an identical 500gb disk, but did find a 2TB just to get things
>>>>> mirrored again.  The bios screen accepted the replacement disk and
>>>>> said it would rebuild in the OS.  mdsync seemed to do its thing but I
>>>>> noticed mdmon process was taking 200% cpu.  I let it go a few days
>>>>> thinking it was just taking longer than normal to sync, then rebooted.
>>>>> It was in a complete failed state and wouldnt boot at all.  After
>>>>> removing the 2TB disk I was able to boot into the OS again.  I just
>>>>> assumed I needed a similar drive size for the second part of the
>>>>> mirror.
>>>>
>>>>
>>>>
>>>> when you talk about a "SATA RAID Controller" and "The bios screen
>>>> accepted
>>>> the replacement disk and said it would rebuild in the OS" this sadly is
>>>> not
>>>> a "linux software raid" at it's own
>>>>
>>>> 197 Current_Pending_Sector  -O--CK   200   200   000    -    9
>>>> 198 Offline_Uncorrectable   ----CK   200   200   000    -    7
>>>>
>>>> i would strongly suggest https://www.gnu.org/software/ddrescue/ and make
>>>> a
>>>> image of that disk because after 39646 Power_On_Hours it's likely that
>>>> the
>>>> remaining disk fails completly in a short time and you could at least
>>>> restore the disk-image with "dd" to a new disk if that happens as well as
>>>> mount it with as loop-device