linux-raid compatable w/ Adaptec AIC-7902?

linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

Hello, All:

I'm trying to install CentOS 3.3 (a.k.a. RedHat Enterprise Linux 3) on a 
Supermicro 5013C-M8 w/ on-board Adaptec AIC-7902 Ultra320 controller 
(connected to two Seagate ST373207LC 73GB hdd's in RAID1 connfiguration) but 
am getting random crashes. (The screen is FILLED with I/O errors and raid1 
messages.)

Is linux-raid compatible with the infamous AIC-7902? Is there any special 
configuration that I need to perform in the Adaptect BIOS?

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 03:38 pm, Eric Pretorious wrote:
>I'm trying to install CentOS 3.3 (a.k.a. RedHat Enterprise Linux 3) on a 
>Supermicro 5013C-M8 w/ on-board Adaptec AIC-7902 Ultra320 controller 
>(connected to two Seagate ST373207LC 73GB hdd's in RAID1 connfiguration) but 
>am getting random crashes. (The screen is FILLED with I/O errors and raid1 
>messages.)

Oops: It's a Supermicro P4SC8 motherboard.

On reboot, the Adaptec BIOS fails to scan the SCSI bus (and, subsequently, the 
system fails to boot). Could this be a hardware failure?

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 03:38 pm, Eric Pretorious wrote:
>I'm trying to install CentOS 3.3 (a.k.a. RedHat Enterprise Linux 3) on a 
>Supermicro 5013C-M8 w/ on-board Adaptec AIC-7902 Ultra320 controller 
>(connected to two Seagate ST373207LC 73GB hdd's in RAID1 connfiguration) but 
>am getting random crashes. (The screen is FILLED with I/O errors and raid1 
>messages.)
>
>Is linux-raid compatible with the infamous AIC-7902? Is there any special 
>configuration that I need to perform in the Adaptect BIOS?

Hm... This sounds fishy:

%> dmesg
...
md: md3: sync done.
md: syncing: RAID array md2
md: minimum _guarenteed_ reconstruction speed: 100KB/sec/disc.
md: using maximum available idle IO bandwidth (but not more than...
md: using 12k window, over a total of 48064 blocks.
md: delaying resync of md0 until md2 has finished resync (they share one or more physical units)
md: delaying resync of md1 until md2 has finished resync (they share one or more physical units)
...

Should I be concerned by the last two lines?

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 03:57 pm, Eric Pretorious wrote:
>On reboot, the Adaptec BIOS fails to scan the SCSI bus (and, subsequently, 
>the system fails to boot). Could this be a hardware failure?

FWIW: I ejected the drives and re-inserted them and the scanning problem went 
away.

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 03:38 pm, Eric Pretorious wrote:
>I'm trying to install CentOS 3.3 (a.k.a. RedHat Enterprise Linux 3) on a 
>Supermicro 5013C-M8 w/ on-board Adaptec AIC-7902 Ultra320 controller 
>(connected to two Seagate ST373207LC 73GB hdd's in RAID1 connfiguration) but 
>am getting random crashes. (The screen is FILLED with I/O errors and raid1 
>messages.)

After rebooting, I saw this message scroll by duing shutdown:

  md4: no spare disk to reconstruct array!  continuing in degraded mode

...so I restarted the system. After the system restarts, the array has the following status...

%> cat /etc/raidtab
  ...
  raiddev	/dev/md4
  raid-level				1
  nr-raid-disks				2
  chuck-size				64k
  perrsistent-superblock	1
  nr-spare-disks			0
	device				/dev/sda7
	raid-disk				0
	device				/dev/sdb7
	raid-disk				1
  ...

%> cat /proc/mdstat
  Personalities: [raid1]
  read_ahead 1024 sectors
  Event: 5
  ...
  md4 : active raid1 sdb7[1]
        50660864 blocks [2/1] [_U]

So it would appear that sda7 has failed. Is that correct?

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 05:38 pm, Eric Pretorious wrote:
>So it would appear that sda7 has failed. Is that correct?

I was in the process of reading this portion of /var/log/messages...

>SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 10000
>I/O error: dev 08:07, sector 2980224
>raid1: disk failure on sda7, disabling device.
>	Operation continuing on 1 devices
>raid1: mirro resync was not fully finished, restarting next time.
>md4: no spare disk to reconstruct array! -- continuing in degraded mode
>md: md_do_sync() got signal ... exiting
>  ...
>md: autorun
>md: considering sdb7 ...
>md:  adding sdb7 ...
>md:  adding sdb7 ...
>md: created md4
>md: bind<sda7,1>
>md: bind<sdb7,2>
>md: running: <sdb7><sda7>
>md: sdb7's event counter: 0000002b
>md: sda7's event counter: 0000002a
>md: superblcokc update time inconsistency -- using the most recent one
>md: freshest: sdb7
>md4: kicking faulty sda7! 
>md: unbind<sda7,1>
>md: export rdev(sda7)
>mf: md4: raid array is not clean -- starting background reconstruction

...when the system choked, again.  Ug!

I expected that a mirror would be more robust than to bring-down the entire 
system when one RAID member fails. Was this assumption incorrect?

-- 
Eric P.,
Truckee, CA

RE: linux-raid compatable w/ Adaptec AIC-7902?

[Guy]

> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org [mailto:linux-raid-
> owner@vger.kernel.org] On Behalf Of Eric Pretorious
> Sent: Wednesday, June 29, 2005 7:56 PM
> To: linux-raid@vger.kernel.org
> Cc: Mike V
> Subject: Re: linux-raid compatable w/ Adaptec AIC-7902?
> 
> On Wednesday 29 June 2005 03:38 pm, Eric Pretorious wrote:
> >I'm trying to install CentOS 3.3 (a.k.a. RedHat Enterprise Linux 3) on a
> >Supermicro 5013C-M8 w/ on-board Adaptec AIC-7902 Ultra320 controller
> >(connected to two Seagate ST373207LC 73GB hdd's in RAID1 connfiguration)
> but
> >am getting random crashes. (The screen is FILLED with I/O errors and
> raid1
> >messages.)
> >
> >Is linux-raid compatible with the infamous AIC-7902? Is there any special
> >configuration that I need to perform in the Adaptect BIOS?
> 
> Hm... This sounds fishy:
> 
> %> dmesg
> ...
> md: md3: sync done.
> md: syncing: RAID array md2
> md: minimum _guarenteed_ reconstruction speed: 100KB/sec/disc.
> md: using maximum available idle IO bandwidth (but not more than...
> md: using 12k window, over a total of 48064 blocks.
> md: delaying resync of md0 until md2 has finished resync (they share one
> or more physical units)
> md: delaying resync of md1 until md2 has finished resync (they share one
> or more physical units)
> ...
> 
> Should I be concerned by the last two lines?
No, this is normal.  Your 3 arrays are on the same disk(s).  In this case,
it is faster to do them 1 at a time.

Guy

> 
> --
> Eric P.,
> Truckee, CA
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Dinesh]

Hello Eric,
>>SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 10000
>>I/O error: dev 08:07, sector 2980224
>>raid1: disk failure on sda7, disabling device.
>> Operation continuing on 1 devices
>>raid1: mirro resync was not fully finished, restarting next time.
>>md4: no spare disk to reconstruct array! -- continuing in degraded mode
>>md: md_do_sync() got signal ... exiting

From messages above it clearly says that sda7 is bad. so remove sda7 and put 
the new disk in and rebuild RAID. RAID will rebuild in background so you 
will feel a little bit slow response from the system. Once it finished RAID, 
system comes back to normal.

> I expected that a mirror would be more robust than to bring-down the 
> entire
> system when one RAID member fails. Was this assumption incorrect?

You don't need bring down system when one RAID member fails. Just replace 
the faulty one with new one.

Thanks
Dinesh 

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Wednesday 29 June 2005 07:10 pm, Guy wrote:
>> md: delaying resync of md0 until md2 has finished resync (they share one
>> or more physical units)
>> md: delaying resync of md1 until md2 has finished resync (they share one
>> or more physical units)
>> ...
>> 
>> Should I be concerned by the last two lines?
>No, this is normal.  Your 3 arrays are on the same disk(s).  In this case,
>it is faster to do them 1 at a time.

Oh - I get it: md can only work on one RAID at a time because md0, md1, & md2 
are all on sda & sdb (i.e., sda is mirrored on sdb). Thanks!

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Turbo Fredriksson]

>>>>> "Eric" == Eric Pretorious <eric@pretorious.net> writes:

    Eric> On Wednesday 29 June 2005 07:10 pm, Guy wrote:
    >>> md: delaying resync of md0 until md2 has finished resync (they
    >>> share one or more physical units) md: delaying resync of md1
    >>> until md2 has finished resync (they share one or more physical
    >>> units) ...
    >>> 
    >>> Should I be concerned by the last two lines?
    >> No, this is normal.  Your 3 arrays are on the same disk(s).  In
    >> this case, it is faster to do them 1 at a time.

    Eric> Oh - I get it: md can only work on one RAID at a time
    Eric> because md0, md1, & md2 are all on sda & sdb (i.e., sda is
    Eric> mirrored on sdb). Thanks!

I'm not DELIBERATELY trying to be rude here, but I want you to find the
answer yourself... :)

WHY are you putting multiple md's on the same disks? Did you intend to
do something like this?

   md0          /boot      sd[ab]1
   md1          /          sd[ab]2
   md2          /usr       sd[ab]3
   md3          /var       sd[ab]4

Just one question (this is the magic :). What do you gain by this? And
why do you do it?
-- 
Marxist Albanian quiche NSA critical FBI Honduras radar president
cracking assassination FSF cryptographic DES Mossad
[See http://www.aclu.org/echelonwatch/index.html for more about this]

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Eric Pretorious]

On Monday 04 July 2005 04:27 am, Turbo Fredriksson wrote:
>WHY are you putting multiple md's on the same disks? Did you intend to
>do something like this?
>
>   md0          /boot      sd[ab]1
>   md1          /          sd[ab]2
>   md2          /usr       sd[ab]3
>   md3          /var       sd[ab]4
>
>Just one question (this is the magic :). What do you gain by this? And
>why do you do it?

I'm just trying to mirror two disks - that's all. Is there a better way?

-- 
Eric P.,
Truckee, CA

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Mathieu Chouquet-Stringer]

eric@pretorious.net (Eric Pretorious) writes:
> I'm just trying to mirror two disks - that's all. Is there a better way?

LVM on top of RAID! ;-)
-- 
Mathieu Chouquet-Stringer
    "Le disparu, si l'on vénère sa mémoire, est plus présent et
                 plus puissant que le vivant".
           -- Antoine de Saint-Exupéry, Citadelle --
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Turbo Fredriksson]

Quoting Eric Pretorious <eric@pretorious.net>:

> On Monday 04 July 2005 04:27 am, Turbo Fredriksson wrote:
>>WHY are you putting multiple md's on the same disks? Did you intend to
>>do something like this?
>>
>>   md0          /boot      sd[ab]1
>>   md1          /          sd[ab]2
>>   md2          /usr       sd[ab]3
>>   md3          /var       sd[ab]4
>>
>>Just one question (this is the magic :). What do you gain by this? And
>>why do you do it?
>
> I'm just trying to mirror two disks - that's all. Is there a better way?

The keyword(s) here is _same disk_.

This is a very complicated way to do it... The idea behind partitioning the
disk(s) this way (many partition, one partition for 'each important FS') is
'to protect the root partition' (or rather 'protect the file systems from
each others failures'). That is, a crash in the /var FS (or that partition)
will not interfere against the /usr FS/partition etc...

This is almost a must to have ANY kind of security against data loss, but
but why do RAID; to protect the FS from hardrive failures...

This two ways of thinking is kinda the same thing, so why do all the
partitioning if one uses RAID?!

Just create ONE partition (plus a swap on each disk, totaling the total amount
of swap you want/need) and set that up as md0 (mounted as /) and that's it!

If a disk/partition crashes, then the RAID 'system' will kick in. If you're
mirroring the disks that is. It won't work if you'r joining disks to get more
space, which is quite pointless in the first place nowadays. Harddisks are so
cheap so just buy a bigger one...


In theory, the way you do it (and the way I used to to it previosly) way will
save you from filesystem (not hardware related!) crashes, but i haven't seen
a filesystem crash in years (ext[23] etc is getting _quite_ stable these days),
but even if the FS crashes the RAID should still kick in to save you data...


This is the way I'm reasoning at least. I might have missed something, and if
I have, I'm interessted in hearing why...
The weak point in my reasoning is if the RAID really saves you from _filesystem_
crashes... I think it does, but I'm not sure...
-- 
KGB Saddam Hussein Nazi toluene cracking NORAD ammunition Ft. Bragg
arrangements Cuba Peking FSF AK-47 Waco, Texas Soviet
[See http://www.aclu.org/echelonwatch/index.html for more about this]

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Joshua Baker-LePain]

On Fri, 8 Jul 2005 at 2:40pm, Turbo Fredriksson wrote

> Quoting Eric Pretorious <eric@pretorious.net>:
> 
> > On Monday 04 July 2005 04:27 am, Turbo Fredriksson wrote:
> >>WHY are you putting multiple md's on the same disks? Did you intend to
> >>do something like this?
> >>
> >>   md0          /boot      sd[ab]1
> >>   md1          /          sd[ab]2
> >>   md2          /usr       sd[ab]3
> >>   md3          /var       sd[ab]4
> >>
> >>Just one question (this is the magic :). What do you gain by this? And
> >>why do you do it?
> >
> > I'm just trying to mirror two disks - that's all. Is there a better way?
> 
> The keyword(s) here is _same disk_.
> 
> This is a very complicated way to do it... The idea behind partitioning the
> disk(s) this way (many partition, one partition for 'each important FS') is
> 'to protect the root partition' (or rather 'protect the file systems from
> each others failures'). That is, a crash in the /var FS (or that partition)
> will not interfere against the /usr FS/partition etc...

There are other reasons to use multiple partitions.  Having /home as a 
separate partition, e.g., allows one to reinstall without copying the data 
off to another machine.  It also prevents a user from filling up the root 
fs, which often has bad consequences.  The same goes for partitions for 
/tmp and/or /var.

> Just create ONE partition (plus a swap on each disk, totaling the total amount
> of swap you want/need) and set that up as md0 (mounted as /) and that's it!

But then if a disk dies, the system is still going to crash, b/c half the 
swap is suddenly gone.  If you want the system to not crash if a disk 
dies, then you must mirror swap as well.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Mark Hahn]

> > >>   md0          /boot      sd[ab]1
> > >>   md1          /          sd[ab]2
> > >>   md2          /usr       sd[ab]3
> > >>   md3          /var       sd[ab]4
...
> There are other reasons to use multiple partitions.  Having /home as a 
> separate partition, e.g., allows one to reinstall without copying the data 
> off to another machine.  It also prevents a user from filling up the root 
> fs, which often has bad consequences.  The same goes for partitions for 
> /tmp and/or /var.

one very serious problem with this approach is that each filesystem
will probably be journaled, which means that it'll spend a lot of time
shuttling the heads between journal and data areas.  if you have activity
on /, /usr and /var at the same time, your heads will do little other 
than rush madly around.  in cases where this can't be avoided (for instance,
using quotas to eliminate the user-fills-partition problem), you can 
often improve things markedly by turning on noatime/nodiratime.

Re: linux-raid compatable w/ Adaptec AIC-7902?

[Tyler]

Mark Hahn wrote:

>>>>>  md0          /boot      sd[ab]1
>>>>>  md1          /          sd[ab]2
>>>>>  md2          /usr       sd[ab]3
>>>>>  md3          /var       sd[ab]4
>>>>>          
>>>>>
>...
>  
>
>>There are other reasons to use multiple partitions.  Having /home as a 
>>separate partition, e.g., allows one to reinstall without copying the data 
>>off to another machine.  It also prevents a user from filling up the root 
>>fs, which often has bad consequences.  The same goes for partitions for 
>>/tmp and/or /var.
>>    
>>
>
>one very serious problem with this approach is that each filesystem
>will probably be journaled, which means that it'll spend a lot of time
>shuttling the heads between journal and data areas.  if you have activity
>on /, /usr and /var at the same time, your heads will do little other 
>than rush madly around.  in cases where this can't be avoided (for instance,
>using quotas to eliminate the user-fills-partition problem), you can 
>often improve things markedly by turning on noatime/nodiratime.
>
>  
>
Also, depending on the file system, you can set reserved space 
percentage for root with tools such as tune2fs.

Tyler.