[parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Dub]

Hi,

1. The document on 
http://www.parisc-linux.org/faq/raidboot-howto.html
is really great. However, there is one exception

Command from FAQ:
mdadm --create /dev/md0 --level 1 --raid-devices=2 missing /dev/sdb2

Command working with curreny woody release:
mdadm -C /dev/md0 --level raid1 --raid-disks 2 missing /dev/sde2

There is no --raid-devices option in my release!

2. 

I have mirrored boot, swap and root devices on pair of physical disks. I took
off one hot-plug disk and returned it back.
root device did not want to resync even after reboot.


md0 : active raid1 sde2[1] sda2[0]
      33728 blocks [2/2] [UU]
      
md1 : active raid1 sde3[1] sda3[0]
      525248 blocks [2/2] [UU]
      
md2 : active raid1 sda5[0]
      4218752 blocks [2/1] [U_]

I fixed it by command

mdadm /dev/md2 -a /dev/sde5

What's wrong???

Kernel was compiled with CONFIG_HOTPLUG=y


Many thanks in advance for your help!

BR,
Dub

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Martin K. Petersen]

>>>>> "Dub" == Dub  <dub@latnet.lv> writes:

Dub> Command from FAQ: mdadm --create /dev/md0 --level 1
Dub> --raid-devices=2 missing /dev/sdb2

Dub> Command working with curreny woody release: mdadm -C /dev/md0
Dub> --level raid1 --raid-disks 2 missing /dev/sde2

Dub> There is no --raid-devices option in my release!

Odd.  I'm running mdadm 1.2.0 and it says:

        allegro:~# mdadm --create --help 2>&1 | grep raid-dev
        Usage:  mdadm --create device -chunk=X --level=Y --raid-devices=Z devices
          --raid-devices= -n : number of active devices in array

Maybe Neil changed it in a later version?


Dub> I have mirrored boot, swap and root devices on pair of physical
Dub> disks. I took off one hot-plug disk and returned it back.  root
Dub> device did not want to resync even after reboot.

Well.  If the disk is no longer part of the metadevice, you manually
have to add it using the mdadm command.

It would be quite dangerous if the RAID code decided to add things to
an array just because a disk is inserted (fine for a storage
appliance, bad idea for a server).

CONFIG_HOTPLUG has nothing to do with it.

-- 
Martin K. Petersen      http://mkp.net/

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Joel Soete]

> It would be quite dangerous if the RAID code decided to add things
> to an array just because a disk is inserted (fine for a storage
> appliance, bad idea for a server).

CONFIG_HOTPLUG has nothing to do with it.

Yes read in <http://www.ibiblio.org/pub/Linux/docs/howto/other-formats/html_single/SCSI-2.4-HOWTO.html#mlproc>

that in Linux 2.4: "The SCSI subsystem does not support hot-plugging of SCSI
devices ..."


Well I have no experience with hpux and those new 'hotplugable' devices but
at least on Sun it doesn't mean that you can remove device(s) like this without
taking care to power it off by software!!!

hth,
    Joel


------------------------------------------------------
Soldes Tiscali ADSL : 27,50 euros/mois jusque fin 2003.
On s'habitue vite à payer son ADSL moins cher!
Plus d'info? Cliquez ici... http://reg.tiscali.be/default.asp?lg=fr 

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Matthew Wilcox]

On Wed, Jul 09, 2003 at 10:29:08AM +0200, Joel Soete wrote:
> Yes read in <http://www.ibiblio.org/pub/Linux/docs/howto/other-formats/html_single/SCSI-2.4-HOWTO.html#mlproc>
> 
> that in Linux 2.4: "The SCSI subsystem does not support hot-plugging of SCSI
> devices ..."

True, it's pretty damn racy.  But if you're careful, it can work.
Let's hope we manage to make it better in 2.6 ... there's a pair of BOFs
in two weeks time at OLS (Hotplug & Linux SCSI) that should address this.

> Well I have no experience with hpux and those new 'hotplugable' devices but
> at least on Sun it doesn't mean that you can remove device(s) like this without
> taking care to power it off by software!!!

Well, no, you're always going to have to quiesce the device.  Otherwise it's
like turning the machine off without halting -- random things are cached and
simply never make it back to disc.

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Joel Soete]

> Well, no, you're always going to have to quiesce the device.  Otherwise
it's
> like turning the machine off without halting -- random things are cached
and
> simply never make it back to disc.

Ah yes, (my bad, i thought but not write) I would have to specify that I
spok about hotplugable I/O device in general (FC, NIC,... with the famous
blue led) and not about disk which we qualify here as 'hotswapable' disk.

Thanks,
    Joel


------------------------------------------------------
Soldes Tiscali ADSL : 27,50 euros/mois jusque fin 2003.
On s'habitue vite à payer son ADSL moins cher!
Plus d'info? Cliquez ici... http://reg.tiscali.be/default.asp?lg=fr 

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Grant Grundler]

On Wed, Jul 09, 2003 at 01:27:00PM +0100, Matthew Wilcox wrote:
> Well, no, you're always going to have to quiesce the device.

Not unless the disk is part of RAID[135] set.
I expect to be able to yank a disk and replace with a new one any time.

grant

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Matthew Wilcox]

On Wed, Jul 09, 2003 at 09:54:51AM -0600, Grant Grundler wrote:
> On Wed, Jul 09, 2003 at 01:27:00PM +0100, Matthew Wilcox wrote:
> > Well, no, you're always going to have to quiesce the device.
> 
> Not unless the disk is part of RAID[135] set.
> I expect to be able to yank a disk and replace with a new one any time.

Indeed.  You are coming to those two BOFs, right?  ;-)

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Grant Grundler]

On Wed, Jul 09, 2003 at 04:58:44PM +0100, Matthew Wilcox wrote:
> Indeed.  You are coming to those two BOFs, right?  ;-)

heh - of course :^)

grant

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Dub]

Hi,

> 
> Odd.  I'm running mdadm 1.2.0 and it says:
> 
>         allegro:~# mdadm --create --help 2>&1 | grep raid-dev
>         Usage:  mdadm --create device -chunk=X --level=Y
> --raid-devices=Z devices
>           --raid-devices= -n : number of active devices in array
> 
> Maybe Neil changed it in a later version?

I downloaded the only  stable version mdadm 0.7.2-2 from ftp.debian.org.   

# mdadm --create --help 2>&1 | grep raid-d
Usage:  mdadm --create device -chunk=X --level=Y --raid-disks=Z devices

I guess it would be nice to mention the difference in the FAQ
 
> Well.  If the disk is no longer part of the metadevice, you manually
> have to add it using the mdadm command.
> 
> It would be quite dangerous if the RAID code decided to add things to
> an array just because a disk is inserted (fine for a storage
> appliance, bad idea for a server).
OK, thank you!


> Yes read in
<http://www.ibiblio.org/pub/Linux/docs/howto/other-formats/html_single/SCSI-2.4-HOWTO.html#mlproc>
> 
> that in Linux 2.4: "The SCSI subsystem does not support hot-plugging
of SCSI
> devices ..."

I just made another test. I pull out 2 mirrored disk (root and var). The system
was running well. However, it seems that md software did not notice absense of
disk until someone tries to write there (I was forced to create a file in /boot
partition so that md understands the failure).

Then I rebooted the system without two disks. PDC told 'warn 80F5' ( Looks Error
executing Entry Init). I do not know why PDC did not try secondary device
automatically, in spite of BOOT ALT worked OK. Looks like PDC was very unhappy
about primary disk and didn't try to try secondary path.

Then Linux didn't find two disks, and reassigned new device names (sda,sdb,...)
according to existing scsi ids. Happily, md software does not care about
symbolic device names and remaps /dev/md  for new disk names according to  hw paths.

Then I inserted the disks and rebooted the system so that the disks appear as
they should. Mirror were resynced manually without problems.

So, I'm not very happy about Linux hot-plug features.

BR,
Dub

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Joel Soete]

> I just made another test. I pull out 2 mirrored disk (root and var). The

> system was running well. However, it seems that md software did not notice
> absense of disk until someone tries to write there (I was forced to create
> a file in /boot partition so that md understands the failure).
> 
> Then I rebooted the system without two disks. PDC told 'warn 80F5' ( Looks
> Error executing Entry Init). I do not know why PDC did not try secondary
> device automatically,
> in spite of BOOT ALT worked OK. Looks like PDC was
> very unhappy about primary disk and didn't try to try secondary path.

do you well set autosearch on ?
Have a look at boot prompt: co bootinfo?

hth,
    Joel


------------------------------------------------------
Soldes Tiscali ADSL : 27,50 euros/mois jusque fin 2003.
On s'habitue vite à payer son ADSL moins cher!
Plus d'info? Cliquez ici... http://reg.tiscali.be/default.asp?lg=fr 

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[dub]

> 
> do you well set autosearch on ?
> Have a look at boot prompt: co bootinfo?
> 
> hth,
>     Joel

Hi,

You are right, autosearch was off. However, setting is on does not help, even
after powering off the system.

BR,
Dub



Configuration Menu: Enter command > bi

BOOT INFORMATION

   Processor         Boot ID
   ---------         -------
       0                  2  
       1                  2  

   Autoboot:             ON
   Autosearch:           ON
   Fastboot:             OFF

   Primary boot path:    10/0.1   (dec)
                         0a/0.1   (hex)
   Alternate boot path:  10/8.1   (dec)
                         0a/08.1   (hex)
   Console path:         10/4/0.0   (dec)
                         0a/04/0.0   (hex)
   Keyboard path:        10/12/7.0   (dec)
                         0a/0c/07.0   (hex)

   LAN Station Address:  080009-da4488

   Thu Jul  10 08:53:10 GMT 2003    (20:03:07:10:08:53:10)

Re: [parisc-linux] Corrections for RAID1 howto and Resyncing after failure

[Joel Soete]

Hi,

> You are right, autosearch was off. However, setting is on does not help,
even
> after powering off the system.
> 
> BR,
> Dub
> 
> 
> 
> Configuration Menu: Enter command > bi
> 
> BOOT INFORMATION
> 
>    Processor         Boot ID
>    ---------         -------
>        0                  2  
>        1                  2  
> 
>    Autoboot:             ON
>    Autosearch:           ON
>    Fastboot:             OFF
> 
>    Primary boot path:    10/0.1   (dec)
>                          0a/0.1   (hex)
>    Alternate boot path:  10/8.1   (dec)
>                          0a/08.1   (hex)

iirc if you launch 'boot alt' and ask to boot 'without interact with ipl'
it boots fine? If yes, i don't see why it doen't work (sorry).
(may be try to switch boot disks to be sure that the behaviour will not follow
the disk?)

Joel


------------------------------------------------------
Soldes Tiscali ADSL : 27,50 euros/mois jusque fin 2003.
On s'habitue vite à payer son ADSL moins cher!
Plus d'info? Cliquez ici... http://reg.tiscali.be/default.asp?lg=fr