Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Solaris 10 with RAID-1 for PE 2850

114 views
Skip to first unread message

Shadow

unread,
Dec 13, 2005, 7:20:16 AM12/13/05
to
What am I doing wrong?


I ve to start with "I'm no disk expert", so please be patient.


My Dell machine order:
- PowerEdge 2850,
- 2 x 146GB SCSI Ultra320,
- PERC 4 Embedded RAID Option.
- C03 MR1, RAID 1 using on-board RAID controller.


(I used the ITU Diskette from Sun for LSI Perc 4e/di drivers,
http://www.sun.com/bigadmin/hcl/data/sol/systems/details/806.html.)
I started the installation of Solaris 10 but I could only see one disk
(c0t0d0) during installation. (?)
I have two Scsi disks, and the idea is that they are set in RAID 1.


Boot path says: /pci@0,0/pci-ide@1f,1/ide@0/sd@0,0:a


Well, I continued the installation and Solaris 10 was finished and it
run well.
Now I wanted to test if RAID 1 was correct, "still".


1. I removed disk 1 and restarted
--------------------------------
The system answered with:
Following SCSI ID's are not responding
Channel-0:0
Please Inspect Enclosure Power, SCSI Cabling, and Drive Connections

and Reboot Your System.


1 Logical Drive(s) found on the host adapter.
0 Physical Drive(s) found on the host adapter.
1 Logical Drive(a) handled by BIOS
0 Physical Drive(s) handled by BIOS
Press <Ctrl><M> to Run Configuration Utility or any other Key to
Continue.


2. I put back disk 1 and restarted
-----------------------------------
Solaris 10 run.


3. I removed disk 0 and restarted
----------------------------------
The system were angry now and I got response:
1 Logical Drive(s) found on the host adapter.
0 Physical Drive(s) found on the host adapter.
1 Logical Drive(s) Degraded
1 Logical Drive(a) handled by BIOS
0 Physical Drive(s) handled by BIOS
Configuration of NVRAM and drives mismatch (Normal mismatch)
User Configuration....


Run View/Add Configuration option of Config Utility
Press A Key to Run Configuration Utility Or <Alt><F10> to Continue.

ALT-F10 ...
No boot device available -
Strike F1 to retry boot, F2 for setup utility


Now it seems that the RAID-1 was not correct, why does Solaris only
find one disk ? where is the other one?
How can I install Solaris and have the RAID 1 from the HW (controller)?

When entering "Utility mode" I can see both disks.


I cleared the configuration and recreated the raid set, and
initialized.
So the system is clean again, and I need to re-install Solaris, but
please advice on how to get RAID 1 ?


Please advice,


Thank you!

tunla

unread,
Dec 13, 2005, 3:02:16 PM12/13/05
to

Shadow wrote:
> What am I doing wrong?

>
>


> 2. I put back disk 1 and restarted
> -----------------------------------
> Solaris 10 run.
>

*******************************************
You need to allow ONE HOUR or MORE
between these two steps so that the disks can be
resyncronised by the RAID controller
before you break the mirror again.
********************************************

Shadow

unread,
Dec 13, 2005, 3:40:08 PM12/13/05
to
The system should use RAID-1, which mean if one disk is not in ( if it
failed ) the system should continue run.

Johan Kielbaey

unread,
Dec 13, 2005, 4:26:33 PM12/13/05
to
Shadow wrote:
> The system should use RAID-1, which mean if one disk is not in ( if it
> failed ) the system should continue run.
>

Indeed. It should keep working if your other disk is still fine.

RAID1 is intended to provide data redundancy (a copy of disk 1 on disk
2). However if you unplug the working disk before it was able to sync
its content on the other disk your system will crash.

More info on raid : http://www.pcguide.com/ref/hdd/perf/raid/index.htm

Rgds,

Johan

tunla

unread,
Dec 13, 2005, 5:34:41 PM12/13/05
to

Shadow wrote:
> The system should use RAID-1, which mean if one disk is not in ( if it
> failed ) the system should continue run.

When you insert disk 1 after the initial Mirror Break - it will
take some to resyncronise , During this time the RAID
controller will have labled the disc 1 as "INVALID" or
RESYNCING or whatever name this particular
RAID controller utilize.

IF you pull the Vaild Active DISC whilst the RESYNCING is IN
progress you are LOST. And your system will CRASH.

//Lars

MyKey

unread,
Dec 14, 2005, 12:06:13 PM12/14/05
to
Being a noob to solaris x86, I ran into broadly similar issues when
trying to raid 1 solaris 10 on a sunfire v20z (2x 73GB HDD). In this
case the Raid controller only registered the disks as 4GB!!

After trying various different methods, I arrived at the sun knowledge
base article below , which seems specific to the server model, but the
same method may be worth a try.

http://sunsolve.sun.com/searchproxy/document.do?assetkey=1-9-78875-1

Below is the main part of the doc. It may help you out. It worked on my
setup, removing (failing) either of the disks still allows the system
to boot and re-sync when disk is re-inserted.

Document Body: Top

While hardware RAID is suppose to be transparent to Solaris[TM] x86,
the setup
of RAID configuration is managed from within the OS.

See instructions below for setting up RAID.

NOTE: if console redirection is enabled and the serial port is used,
refer
to the "Sun Fire V20z Server User Guide" (817-5248-11) for the correct
escape sequences (2.3 Escape Sequences for Remote Console Terminal).

1. Install the OS on the primary drive of the system.

2. Once this is complete, install update 116667-02 or later, as this
provides
the utility, raidctl, needed to setup the second drive for hardware
RAID using
the LSI 1020 controller.

3. With the OS booted up, login as root, and run the following command:

raidctl -c disk1 disk2

where:

disk1 = canonical name of the primary internal disk
disk2 = canonical name of the second internal disk

The disks must be specified by its respective canonical name, such as,
c0t0d0.

When you create a a RAID volume, the RAID volume assumes the
identity of
the first target in the disk pair (disk1). The second target (disk2)

disappears from the system. Therefore, the RAID volume appears as one
disk.

For additional information on the different usages of the command,
raidctl,
please see the man page by typing at the command prompt:

man raidctl

For information on setting up hardware RAID for linux, please see
document ID
76805.

Shadow

unread,
Dec 14, 2005, 4:49:49 PM12/14/05
to
Thank you Lars, Johan and Michael for all your help and guide!

I've now re-installed Solaris 10 again and the RAID-1 is working, and
now I m careful for not pull both disks...

I wonder if there is any way to see the disk health without going to
the BIOS ?

raidctl does not tell anything.
dmesg and var/adm/messages are quiet when I pull a disk!


I mean, it sounds dangerous if Solaris does not report anything.. what
happen if one disk fails and Solaris doesnt report.. and after a while
the other disk fails too.
I will have a system with two broken disks.. no pre-warnings!!

Thank you guys!

Johan Kielbaey

unread,
Dec 14, 2005, 5:52:09 PM12/14/05
to
Shadow wrote:
> Thank you Lars, Johan and Michael for all your help and guide!
>
> I've now re-installed Solaris 10 again and the RAID-1 is working, and
> now I m careful for not pull both disks...
>
> I wonder if there is any way to see the disk health without going to
> the BIOS ?

I googled a bit about this, but was not able to find anything useful.
Unfortunately Dell doesn't offer a driver for Perc 4 for Solaris x86. To
be honest my experience of Dell's support on a non-windows OS is very poor.

I did find that Solaris Express 10/2005 contains a lsimega driver which
supports perc4 RAID controller. However this won't help you much since
you're already able to install Solaris with some kind of driver that is
supplied by Solaris.

The nice thing about a hardware raid controller is that the OS doesn't
see and doesn't have to care about raid or physical disks. For all the
OS knows the disk it sees is a physical disk. The bad part of this is
that without the appropriate tool you feel like a sitting duck, waiting
till both disks are down and you're in problems. Dell provides a raid
manager for all possible Windows-versions and for very limited Linux
version, but no Solaris.

Perhaps you could check out /proc. There might be some indications of a
rebuilding progress.

If the status of each individual drive is important to monitor you can
always use Volume Manager >;-) It has proven to work great and
performance loss is limited (except during rebuild) on sparc.

>
> raidctl does not tell anything.
> dmesg and var/adm/messages are quiet when I pull a disk!
>
>
> I mean, it sounds dangerous if Solaris does not report anything.. what
> happen if one disk fails and Solaris doesnt report.. and after a while
> the other disk fails too.
> I will have a system with two broken disks.. no pre-warnings!!

Since the hardware controller keeps the physical details hidden from
Solaris you will not see a disk fail.

>
> Thank you guys!
>

Johan

Shadow

unread,
Dec 15, 2005, 3:56:26 AM12/15/05
to
You only shed light Johan,

great thanks!!


Well, I ve sent a message to Dell forum. We'll see what they say.

But about SVM, it is nice I ve used it alot and this is first time Im
using Hardware Raid (and first time I use x86 - in production).

But I can not use SVM for this machine, while I ve tested this
scenarios in my BIOS:

1.
Embeded RAID controller Off
Channel A Off
Channel B Off
NO disks could be found during Solaris installation

2.
Embeded RAID controller SCSI
Channel A SCSI
Channel B SCSI
No RAID used, just the 2 SCSI disks.
Solaris installation hang at "Configuring devices"
Booting kernel/unix...
SunOS Release...
Copyright ...
Use is subject...
Configuring device. <--- here it hang in hours

3.
Embeded RAID controller RAID enabled
Channel A RAID
Channel B SCSI
The only one that works, but then I can not use SVM as I use the
embeded hardware RAID.


Thank you!

0 new messages