when --add after a fail?

when --add after a fail?

[Sandro Dentella]

Hi all,

   i've been using raid1 sisnce quite a loto of time. Sporadiccaly an array
   fails a disk and in many situations I can just pull the device into the
   array with

      mdadm /dev/mdN --add failded_device

   I've never really understood what is the magic that resurrexes it. I
   thought something related to relocation of bad blocks, but I'm  not at
   all aware of what happens "there"...

   Is that a correct trial to do?

   Now I have a device that throuws an error:

srv-ornago:/tmp# mdadm /dev/md2 --add /dev/hdc6 
mdadm: hot add failed for /dev/hdc6: Invalid argument

the kernel complains:
Mar  1 09:34:29 srv-ornago kernel: md: could not bd_claim hdc6.
Mar  1 09:34:29 srv-ornago kernel: md: error, md_import_device() returned -16

   is this enought to say the disk (pretty new, 6 months) is to be changed?
   which checks should I do.

Thanks in advance
sandro
*:-)


srv-ornago:/tmp# mdadm  --detail /dev/md2
/dev/md2:
        Version : 00.90.01
  Creation Time : Mon Apr 29 02:29:18 2002
     Raid Level : raid1
     Array Size : 5116544 (4.88 GiB 5.24 GB)
    Device Size : 5116544 (4.88 GiB 5.24 GB)
   Raid Devices : 2
  Total Devices : 2
Preferred Minor : 2
    Persistence : Superblock is persistent

    Update Time : Wed Mar  1 10:18:15 2006
          State : clean, degraded
 Active Devices : 1
Working Devices : 1
 Failed Devices : 1
  Spare Devices : 0

           UUID : c203673d:0444f961:50e95acb:fb5b89ba
         Events : 0.1392171

    Number   Major   Minor   RaidDevice State
       0       3        6        0      active sync   /dev/.static/dev/ide/host0/bus0/target0/lun0/part6
       1       0        0        -      removed

       2      22        6        -      faulty   /dev/.static/dev/ide/host0/bus1/target0/lun0/part6

   

-- 
Sandro Dentella  *:-)
e-mail: sandro@e-den.it 
http://www.tksql.org                    TkSQL Home page - My GPL work

a user-raid list for non gurus?

[Sandro Dentella]

Hi all,

  even thought I'm subscribed to this list since quite a lot of time, I only
  can grab a little purcentage of the discussion due to technical gap.
  
  When it happens that I ask something to the list I normally get unanswered
  even on problems that in my opinion shouldn't be considered really
  specific to my setup. Here just the last 2 questions:

http://marc.theaimsgroup.com/?l=linux-raid&m=114185900020437&w=2
http://marc.theaimsgroup.com/?l=linux-raid&m=114120476328121&w=2
  
  I'd like to know if there is a users-linux-raid where these kind of
  questions get answered, if I should just insist on this list, or if I
  missed some document (possibly a complete howto) that already gives 
  thorought info on how to cope with troubles, or a wiki.
  
  I feel like linux-raid is missing a good howto. The documentation you can
  find is often outdated or incomplete and it only saldomly helped me to get
  out of problems. 

  The sad part of all this is that with raid systems you are not normally in
  the mood/position to test things. You would much more prefere to arrive at
  the crash already with the knowledge on how to cope with it... but the
  crash is never as the one you simulated failing the device...

Thanks for your attention and you job anyhow
sandro
*:-)


PS: In case anybody feeels like writing a complete howto I could
    help... with the index ;-) of what a system admin -not a raid guru-
    would like to read to understant raid (not just to setup)


-- 
Sandro Dentella  *:-)
e-mail: sandro@e-den.it 
http://www.tksql.org                    TkSQL Home page - My GPL work

Re: a user-raid list for non gurus?

[Neil Brown]

On Friday March 10, sandro@e-den.it wrote:
> Hi all,
> 
>   even thought I'm subscribed to this list since quite a lot of time, I only
>   can grab a little purcentage of the discussion due to technical gap.
>   
>   When it happens that I ask something to the list I normally get unanswered
>   even on problems that in my opinion shouldn't be considered really
>   specific to my setup. Here just the last 2 questions:
> 
> http://marc.theaimsgroup.com/?l=linux-raid&m=114185900020437&w=2
> http://marc.theaimsgroup.com/?l=linux-raid&m=114120476328121&w=2

Maybe you just have to ask again.  It worked this time....

>   
>   I'd like to know if there is a users-linux-raid where these kind of
>   questions get answered, if I should just insist on this list, or if I
>   missed some document (possibly a complete howto) that already gives 
>   thorought info on how to cope with troubles, or a wiki.

No, this list is the best place to ask.  Questions usually get
answered, but not always. People are busy etc.  Feel free to ask a
second time if you haven't had an answer.

>   
>   I feel like linux-raid is missing a good howto. The documentation you can
>   find is often outdated or incomplete and it only saldomly helped me to get
>   out of problems. 
> 
>   The sad part of all this is that with raid systems you are not normally in
>   the mood/position to test things. You would much more prefere to arrive at
>   the crash already with the knowledge on how to cope with it... but the
>   crash is never as the one you simulated failing the device...
> 
> Thanks for your attention and you job anyhow
> sandro
> *:-)
> 
> 
> PS: In case anybody feeels like writing a complete howto I could
>     help... with the index ;-) of what a system admin -not a raid guru-
>     would like to read to understant raid (not just to setup)

To have a good howto, we need someone to step forward to write it.
They don't necessarily need to know all the answers - they can ask
here - but it would help if they know some good questions, and are
reasonably good at technical writing.

NeilBrown

Re: a user-raid list for non gurus?

[Sandro Dentella]

> > http://marc.theaimsgroup.com/?l=linux-raid&m=114185900020437&w=2
> > http://marc.theaimsgroup.com/?l=linux-raid&m=114120476328121&w=2
> 
> Maybe you just have to ask again.  It worked this time....

thanks for the encouragement and for the nice reply. 

> To have a good howto, we need someone to step forward to write it.
> They don't necessarily need to know all the answers - they can ask
> here - but it would help if they know some good questions, and are
> reasonably good at technical writing.

wouldn't you think that a good wiki with a structured skeleton could help
knolwdged people give their contribution without asking too much time?
Should people feel this is any interesting I could try to find the time to
help in this (on the 'questions' side rather that on the 'ansers' one ;-)

sandro
*:-)


-- 
Sandro Dentella  *:-)
e-mail: sandro@e-den.it 
http://www.tksql.org                    TkSQL Home page - My GPL work

Re: a user-raid list for non gurus?

[Bill Davidsen]

Sandro Dentella wrote:

>Hi all,
>
>  even thought I'm subscribed to this list since quite a lot of time, I only
>  can grab a little purcentage of the discussion due to technical gap.
>  
>  When it happens that I ask something to the list I normally get unanswered
>  even on problems that in my opinion shouldn't be considered really
>  specific to my setup. Here just the last 2 questions:
>
>http://marc.theaimsgroup.com/?l=linux-raid&m=114185900020437&w=2
>http://marc.theaimsgroup.com/?l=linux-raid&m=114120476328121&w=2
>  
>  I'd like to know if there is a users-linux-raid where these kind of
>  questions get answered, if I should just insist on this list, or if I
>  missed some document (possibly a complete howto) that already gives 
>  thorought info on how to cope with troubles, or a wiki.
>  
>

The problem is that with a user list, the people who are best able to 
answer your questions are unlikely to see them. While many people have 
good intentions, in practice we lack TIME to read yet another list, 
particularly those where the chance of finding something useful to an 
experienced user are small.

>  
>  I feel like linux-raid is missing a good howto. The documentation you can
>  find is often outdated or incomplete and it only saldomly helped me to get
>  out of problems. 
>  
>

Right now RAID is moving so fast that it would be out of date, and 
that's less helpful than nothing. The mdadm man page is the howto, and 
the only thing remotely able to provide information on exactly what you 
have installed today.

>  The sad part of all this is that with raid systems you are not normally in
>  the mood/position to test things. You would much more prefere to arrive at
>  the crash already with the knowledge on how to cope with it... but the
>  crash is never as the one you simulated failing the device...
>
>Thanks for your attention and you job anyhow
>sandro
>*:-)
>
>
>PS: In case anybody feeels like writing a complete howto I could
>    help... with the index ;-) of what a system admin -not a raid guru-
>    would like to read to understant raid (not just to setup)
>  
>
I agree that people don't feel like messing with their RAID when they have it working! I would love to change my setup, but there is no funding for a backup system to make backing up several TB convenient, and no time to spend doing it the hard way.

-- 
bill davidsen <davidsen@tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979

Re: when --add after a fail?

[Neil Brown]

On Wednesday March 1, sandro@e-den.it wrote:
> Hi all,
> 
>    i've been using raid1 sisnce quite a loto of time. Sporadiccaly an array
>    fails a disk and in many situations I can just pull the device into the
>    array with
> 
>       mdadm /dev/mdN --add failded_device

If a drive fails, you should try to understand *why* it failed before
simply adding it back to the array.  If it was a transient read error,
then it is fairly safe to add it back, though more recent kernels will
not kick a drive in this situation.
If it was a write error, then the reconstruction will fail.
If it was a cabling or hardware error, then you really need to get it
fixed.


> 
>    I've never really understood what is the magic that resurrexes it. I
>    thought something related to relocation of bad blocks, but I'm  not at
>    all aware of what happens "there"...

md will simply copy all the data from the 'good' drive to the 'bad'
drive.  If this work, life continues happily.  If not....


> 
>    Is that a correct trial to do?
> 
>    Now I have a device that throuws an error:
> 
> srv-ornago:/tmp# mdadm /dev/md2 --add /dev/hdc6 
> mdadm: hot add failed for /dev/hdc6: Invalid argument
> 
> the kernel complains:
> Mar  1 09:34:29 srv-ornago kernel: md: could not bd_claim hdc6.
> Mar  1 09:34:29 srv-ornago kernel: md: error, md_import_device()
> returned -16

This is telling you that the device (hdc6) is in use by something
else.
Is it mentioned in 'cat /proc/mdstat' ? It some partition on it
mounted?

NeilBrown