Topic: HDD Configuration Question -> Is this possible?

[ngomes]

Hi all,

I am trying to setup a SME7 Server with this configuration:

## Hardware and BIOS Settings ##
- Asus Vintage PH1 (1 ATA/100 channel + 4 SATA channels)
- 1 x Seagate HDD ST3400014A ATA/100 40GB as primary master.
- 1 x Samsung CD-ROM SC-152 ATA as primary slave.
- 1 x Segate HDD ST3120827AS SATA 120GB as secondary master.
- Onboard IDE Operate Mode = Enhanced Mode.
- Enhanced Mode Support On = SATA Mode.

The final hardware configuration will have 2 more 120GB SATA disks to build a software RAID5 to storage business and users files. Meanwhile I do not have these 2 additional disks.

## My Testing Procedures ##
1) After performing a CD install with only the ATA disk plugged into the mainboard, I connected the SATA disk and it gave me a kernel panic error.
2) After performing a CD (re)install with both ATA and SATA disks plugged into the mainboard it gave me no errors but SME7 is now (as I am writing) building a RAID1 between the ATA and SATA disks (using the 40GB space of the ATA disk and the 2 partitions created inside it on install).

## What I want to do ##
1) I want to run SME7 on only 1 disk: the 40GB ATA disk.
2) I want to build a software RAID5 using 3 x 120GB SATA disks to storage data files.

## My questions ##
1) Can I do this install (SME7 on 1 disk + Data on a RAID5)?
2) If so, can someone give me some directions?

Thanks in advanced,
Nuno

[ngomes]

I am thinking on reset the current 40GB raid1 (with 2 disks) to a raid1 one-way (1 disk, raid1 degraded) running only on the ATA disk, and make a new partition again on my 120GB sata disk.

What can you say about it?

Thanks in advanced,
Nuno

[ngomes]

Hi again,

To simplify what I want to do with SME7pre3:

# SME7 installation on a 1 (one) Seagate HDD ST3400014A ATA/100 40GB.
# Data files on 3 Seagate HDD ST3120827AS SATA 120GB built as RAID5 after the SME installation.

I don't know from where to start since SME7 doesn't give me a partition tool when I am performing the installation.

It seems to me that anaconda detects how many disks I have and make a decision to build automatically a RAID based on the number of disks connected to motherboard during the installation.

Regards,
Nuno

[gordonr]

I don't know from where to start since SME7 doesn't give me a partition tool when I am performing the installation.

It seems to me that anaconda detects how many disks I have and make a decision to build automatically a RAID based on the number of disks connected to motherboard during the installation.

The supported disk layouts for SME7 are shown here:

http://no.longer.valid/phpwiki/index.php/SME%207%20Features

You have at least three choices:

- Drop the 40GB disk out of the mix and just use a RAID5 install
- Disconnect the disks for the RAID5 and manually build the RAID5 device after the install
- Create a custom kickstart file

[ngomes]

Thanks Gordon for your reply.

You have at least three choices:
- Drop the 40GB disk out of the mix and just use a RAID5 install
- Disconnect the disks for the RAID5 and manually build the RAID5 device after the install
- Create a custom kickstart file

I am trying a different install. Here is what I did today:

1) SME7 cd-rom install using only the 40GB ata disk attached (and the cd-rom reader).
2) Halt the server.
3) Replace the cd-rom reader by another 40GB ata disk.
4) Reboot the server.
5) fdisk /dev/hda to print the partition table down.
6) fdisk /dev/hdb to create 2 partitions with exact same sizes and types as in 5)
7) mdadm --manage /dev/md1 --add /dev/hdb1 to manually add a 2nd partition to the md1 array (for /boot).
mdadm --manage /dev/md2 --add /dev/hdb2 to manually add a 2nd partition to the md2 array (for /)
9) Wait until md1 and md2 are rebuilt (mdadm --detail /dev/md1 and mdadm --detail /dev/md2)
10) Go to server-manager -> administration -> view log files -> selected raidmonitor/current to see if everything went well. The log file was:

Code: [Select]


2006-02-22 17:18:02.867111500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 17:24:35.817144500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 18:17:27.139935500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 18:26:38.339954500 This appears to be a single disk system - disabling raid monitoring

OK. Something is wrong with this log file. I think I have already a RAID1 running but continue to next step anyway:

11) Halt the server
12) Just for testing, add the 120GB sata disk (as sda)
13) Reboot the server
14) SME Server detects the new sata disk -> Fine!
15) fdisk /dev/sda and make a Linux partition
16) mkfs.ext3 /dev/sda1 -c to check for badblocks (currently doing this)

What went wrong with my manual raid1 build to not see "good" logs in 10)?

Regards,
Nuno

[gordonr]

5) fdisk /dev/hda to print the partition table down.
6) fdisk /dev/hdb to create 2 partitions with exact same sizes and types as in 5)
7) mdadm --manage /dev/md1 --add /dev/hdb1 to manually add a 2nd partition to the md1 array (for /boot).
mdadm --manage /dev/md2 --add /dev/hdb2 to manually add a 2nd partition to the md2 array (for /)
9) Wait until md1 and md2 are rebuilt (mdadm --detail /dev/md1 and mdadm --detail /dev/md2)

Or simply use "Manage Disk Redundancy" from the console to add the second disk, and report a bug to the bug tracker if it doesn't work.

[ngomes]

Thanks again Gordon,

Here is what I get from my system:

1) While (re)booting the machine I get this:

Code: [Select]

Setting up Logical Volume Management [OK]
mdadm: only specify super-minor once, super-minor=2 ignored
mdadm: only specify super-minor once, super-minor=1 ignored
No software RAID disks
Setting up Logical Volume Management [OK]

2) Using "Manage Disk Redundancy" from the console I get this:

Code: [Select]

Current RAID status:

Personalities: [raid1]
md2: active raid1 hdb2[1] hda2[0]
         38973568 blocks [2/2] [UU]

md1: active raid1 hdb1[1] hda1[0]
         104320 blocks [2/2] [UU]

unused devices: <none>

All RAID devices are in clean state.

2) From server-manager -> administration -> view log files -> raidmonitor/current log file:

Code: [Select]

2006-02-22 17:18:02.867111500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 17:24:35.817144500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 18:17:27.139935500 This appears to be a single disk system - disabling raid monitoring
2006-02-22 18:26:38.339954500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 09:55:43.451300500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 10:02:01.035293500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 10:06:32.069915500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 10:08:25.039232500 This appears to be a single disk system - disabling raid monitoring

Something is wrong with this log file I suppose. Since I didn't know it was possible to add a second disk from the text console option "Manage Disk Redundancy", perhaps it is better to restart again and see how it goes, isn't it?

Thanks in advanced,
Nuno

[gordonr]

\
Something is wrong with this log file I suppose. Since I didn't know it was possible to add a second disk from the text console option "Manage Disk Redundancy", perhaps it is better to restart again and see how it goes, isn't it?

The raid monitoring doesn't know you've added the second disk. If you do it through the console it will, but please raise this as a bug. We should properly detect a manually added second half of a mirror.

However, the console is preferable - it adds the mirror, configures GRUB to boot from either disk and leaves information which allows the raid monitor to detect the new two-disk RAID. If you do it manually and don't do this, you may end up with a mirror which won't boot when one of the disks is removed.

[ngomes]

Thanks again Gordon,

The raid monitoring doesn't know you've added the second disk. If you do it through the console it will, but please raise this as a bug. We should properly detect a manually added second half of a mirror.

However, the console is preferable - it adds the mirror, configures GRUB to boot from either disk and leaves information which allows the raid monitor to detect the new two-disk RAID. If you do it manually and don't do this, you may end up with a mirror which won't boot when one of the disks is removed.

Here is what I will do today:
1) put my 2 raid1 devices in degraded mode by removing the 2nd disk (and its partitions) from md1 and md2.
2) add the mirrors through "Manage Disk Redundancy" console option.
3) report it in here.

Note: That's because I don't have much experience with grub and mdadm I am a "little" lost on this setup. However, after searching the internet for a while, I discovered this document on an australian mailing list about running raid systems with grub, if someone is interested:
# GRUB + RAID howto

About the bug, I will raise it after my conclusions. Perhaps it is not bug at all as you explained but rather lack of knowledge from me (allways learning...)

Nuno

[ngomes]

Ok, I restarted my SME7 installation once more from the beginning   following Gordon's instructions (through the Manage Disk Redundancy" console option):

Here are the important info:

#1# On booting (same as before):

Code: [Select]

Setting up Logical Volume Management [OK]
mdadm: only specify super-minor once, super-minor=2 ignored
mdadm: only specify super-minor once, super-minor=1 ignored
No software RAID disks
Setting up Logical Volume Management [OK]

#2# On "Manage Disk Redundancy" console option (same as before):

Code: [Select]

Current RAID status:

Personalities: [raid1]
md2: active raid1 hdb2[1] hda2[0]
         38973568 blocks [2/2] [UU]

md1: active raid1 hdb1[1] hda1[0]
         104320 blocks [2/2] [UU]

unused devices: <none>

All RAID devices are in clean state.

#3# From server-manager -> administration -> view log files -> raidmonitor/current log file:

Code: [Select]


2006-02-23 15:41:26.549507500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 15:45:58.700644500 This appears to be a single disk system - disabling raid monitoring
2006-02-23 15:51:50.662822500 mdadm: only specify super-minor once, super-minor=2 ignored.
2006-02-23 15:51:50.678905500 mdadm: only specify super-minor once, super-minor=1 ignored.
2006-02-23 15:51:51.358892500 Event: DegradedArray, Device: /dev/md1, Member:
2006-02-23 15:51:52.124277500 Event: SparesMissing, Device: /dev/md1, Member:
2006-02-23 15:51:52.459560500 Event: DegradedArray, Device: /dev/md2, Member:
2006-02-23 15:51:52.752825500 Event: SparesMissing, Device: /dev/md2, Member:
2006-02-23 16:02:25.041251500 mdadm: only specify super-minor once, super-minor=2 ignored.
2006-02-23 16:02:25.060556500 mdadm: only specify super-minor once, super-minor=1 ignored.
2006-02-23 16:02:25.649994500 Event: DegradedArray, Device: /dev/md1, Member:
2006-02-23 16:02:26.375267500 Event: SparesMissing, Device: /dev/md1, Member:
2006-02-23 16:02:26.670833500 Event: DegradedArray, Device: /dev/md2, Member:
2006-02-23 16:02:27.042859500 Event: SparesMissing, Device: /dev/md2, Member:
2006-02-23 16:04:27.250168500 Event: SpareActive, Device: /dev/md1, Member: /dev/hdb1
2006-02-23 16:04:27.548112500 Event: RebuildStarted, Device: /dev/md2, Member:
2006-02-23 16:09:27.766708500 Event: Rebuild20, Device: /dev/md2, Member:
2006-02-23 16:14:28.019592500 Event: Rebuild40, Device: /dev/md2, Member:
2006-02-23 16:19:28.282992500 Event: Rebuild60, Device: /dev/md2, Member:
2006-02-23 16:24:28.543569500 Event: Rebuild80, Device: /dev/md2, Member:
2006-02-23 16:31:28.814715500 Event: RebuildFinished, Device: /dev/md2, Member:
2006-02-23 16:31:29.080463500 Event: SpareActive, Device: /dev/md2, Member: /dev/hdb2
2006-02-23 16:48:40.106849500 mdadm: only specify super-minor once, super-minor=2 ignored.
2006-02-23 16:48:40.106853500 mdadm: only specify super-minor once, super-minor=1 ignored.
2006-02-23 16:48:40.971651500 Event: SparesMissing, Device: /dev/md1, Member:
2006-02-23 16:48:41.373098500 Event: SparesMissing, Device: /dev/md2, Member:
2006-02-23 17:00:20.170413500 mdadm: only specify super-minor once, super-minor=2 ignored.
2006-02-23 17:00:20.170416500 mdadm: only specify super-minor once, super-minor=1 ignored.
2006-02-23 17:00:20.788688500 Event: SparesMissing, Device: /dev/md1, Member:
2006-02-23 17:00:21.362136500 Event: SparesMissing, Device: /dev/md2, Member:

I don't know what idoes it mean the "Event: SparesMissing, Device: /dev/md1, Member:" text message log so I don't know if everything is fine now. Is it?

Nuno

[ngomes]

However, the console is preferable - it adds the mirror, configures GRUB to boot from either disk and leaves information which allows the raid monitor to detect the new two-disk RAID. If you do it manually and don't do this, you may end up with a mirror which won't boot when one of the disks is removed.

You are right: I have tested my raid1 built through the console booting from one disk only, rebuilding the raid1 through the same console and booting again from the other disk and everything went all

About the bug: Do you want me to raise a bug about my previous manual raid1 build?

Regards,
Nuno