AlphaServer DS20 StorageWork problem

[Marcelo Farias]

Hi friends... I need your help.
I have a AlphaServer DS20 runnig with TRU64, 2 SCSI Disk...
In the StorageWork drive shelf I have the scsi disks with a green led and a orange led.
Since some days ago, the orange led of the scsi disk 1 still on... what is that? what mean the orange led "ON" all the time?

[Larry De Clay]

Hi Marcelo,
May be you have a disk failure, I don´t know if you have a mirror pair, if not you have to notice that one disk is offline and the information was gone.
If you have some mirror functionality (hardware or software) you have to replace the disk failed and reconstruct the mirror.

Bye,
Larry

[AwadheshPandey]

orange disk blinks means some thing wrong with the disk on which it glows, try to delete the disk and re-insert it on bay, if problem is continue then disk need to replace as soon as posible.

Awadhesh

[AwadheshPandey]

orange led means disk failed, first delete this disk from storage and remove the disk from bay, re-insert and add it in storage, if problem continue then disk need to replace as soon as possible,

Awadhesh

[Michael Schulte zur Sur]

Hi,

as the others said, the disk is gone or about to go bad.
You need to know, which controller serves the disk and whether you use LSM or hardware mirroring.
You have to replace the disk anyway.
What is the os version?

greetings,

Michael

[Marcelo Farias]

hi
First, thanks for all the replies.
I have a new question or problem:
I don´t have a mirror system or backup system for my wrong disk.
I´m using Digital Unix 4.0, with Informix Database Server. I´m using a tape backup for DB only.
My question is: If I replace my wrong disk, can I clone it (in the new disk)? What steps or software I have to use?
I don´t want to lost the data in the wrong disk.
The wrong disk only have data. It not is a system or boot disk.
Thanks again
Marcelo

[Vladimir Fabecic]

Marcelo
Please answer Michael's question. We need more informations to help you.
Is you system still working? Is DB still running OK?
If so, maybe you have RAID controller or LSM.
Tell us more details.

[Ivan Ferreira]

You should use vdump/vrestore to save and restore the data on the new disk.

The procedure can be:

use vdump to save the data to tape:

vdump -0uf /dev/rmt0h /mount_point

Replace the disk, configure it and mount it on the same place.

Restore from tape

vrestore -xf /dev/rmt0h -D /mount_point


Or install the new disk and mount it on a temporary location, then use:

vdump -0f - /mount_point_of_failed_disk | vrestore -xf - -D /mount_point_of_new_disk

After the copy finishes, umount the old, remove the disk, and mount the new in the final location.

[Michael Schulte zur Sur]

Marcelo,

you will have to replace the defective disk anyway. Is there any green light activity at all? Some controller do an automatic rebuild of a raid drive if the failed drive is replaced. For LSM you would have to do some other steps. So please tell us what you use.
Post
volprint -Aht
then we can see if you use LSM.

Michael

[Marcelo Farias]

Hi
How I know if I have RAID or LSM?
What is LSM?
The system and DB works fine.

Thanks again friends

[Michael Schulte zur Sur]

LSM is the Logical Storage Manager and is a software solution for mirroring disks.
Pleast post the output of
volprint -Aht
Do you know sys_check?
If it is on your system you can run
sys_check > /tmp/config.html
and post it as attachment.

thanks,

Michael

[Marcelo Farias]

Well.. I´m following the steps..
I try to run volinfo -Aht and no exist.
Running sys_check, I have the html file (1.4 Mb)
What sections you need to view? I can paste it and send it you.

Read this:
<=[ ToC ]=> Logical Storage Manager

LSM vold is not running.

Note: LSM is not being used on the system.
Compaq recommends that customers with Tru64 UNIX midrange and large servers use LSM in order to improve storage subsystem management and availability. LSM also provides enhanced I/O statistics that can facilitate I/O tuning.


System Census for axpmedicina

--------------------------------------------------------------------------------

Run on Tue Nov 15 13:01:38 GMT-0300 2005 Script sys_check Version: 112.0
/vmunix is the current running kernel
Digital UNIX V4.0F (Rev. 1229); Fri Oct 6 10:29:02 GMT-0300 2000

System Type: DEC6600
Number of CPUs: 2 Type: EV6 Speed: 500 Mhz Cache: 4.0 MB Memory size: 512 MB
Security Information Excluded
Warnings and Data Included
System recovery data stored in /var/recovery/sys_check

Thanks for all

[Michael Schulte zur Sur]

Please compress it with gzip and post it here as an attachment.
I can point you to the place that has the information I need.
At least we know that you do not use LSM ans since the system does not miss anything important we can assume that you have some form of hardware raid.
I want to have a look at this html file to see what raid controllers you have.

Michael

[Marcelo Farias]

Ok. Here is the config.html for you!!!
Thank... and very luck!

[Michael Schulte zur Sur]

Hi,

upon inspecting your syscheck file I must say you have a strange configuration.
Please confirm if you have just 3 disks in the whole system. It looks that you are running non mirrored disks.
There are advfs partitions on rz9 but I can't see any domain they are used by.
First of all we must locate the disk with the orange light.
Use
dd if=/dev/rrz8c of=/dev/null bs=1024k
dd if=/dev/rrz9c of=/dev/null bs=1024k
dd if=/dev/rrz10c of=/dev/null bs=1024k
Then you know the disk accessed.
Please post the disk name.
You should then backup and replace the faulty disk. I hope it is not the disk with the raw partitions.

greetings,

Michael

[Marcelo Farias]

Yes friend, I have 3 disk in the server.
Rz8 and Rz9 were included in the server when we buy it.
Rz10 is a new disk 8Gb (used for backups)
Test result:
Rz9 have the orange light "ON".

Next step?

Thanks for all

[Michael Schulte zur Sur]

Ok,

so the disk in question is not in use, right?
If you do not need the data on it, you should shutdown the machine, replace the disk start the machine and
scu scan edt
scu show edt
disklabel -wr -t advfs /dev/rrz9a

Michael

ps. I hope you are not using this as a production machine.

[Johan Brusche]

Marcelo,

Are you sure the Quantum disk is qualified for the DS20 and his SCSI-interface ?

There is nothing in the UERF output that points towards a problem with any of the disk.

And yes, rz9 is used by the Informix DB in raw mode. (see ls -l /dev/* | grep informix )

crw-rw---- 2 informix informix 8,17411 Aug 21 2002 ol_chk1
crw-rw---- 2 informix informix 8,17412 Aug 16 2002 ol_chk2
crw-rw---- 2 informix informix 8,17413 Feb 11 2002 ol_chk3
crw-rw---- 2 informix informix 8,17414 Feb 11 2002 ol_chk4

crw-rw---- 2 informix informix 8,17411 Aug 21 2002 rrz9d
crw-rw---- 2 informix informix 8,17412 Aug 16 2002 rrz9e
crw-rw---- 2 informix informix 8,17413 Feb 11 2002 rrz9f
crw-rw---- 2 informix informix 8,17414 Feb 11 2002 rrz9g

Rgds,
__ Johan./

[Michael Schulte zur Sur]

Johan,

thanks for bringing that to attention.
May be the disk is not yet totally failed?

Michael

[Marcelo Farias]

Johan and Michael
Thanks for all

The situation is:
- Rz9 is used by Informix DB in raw mode (like Johan said) I verify it using ls -l /dev/* | grep informix
- The Quantum disk are installed and working fine... we don´t have any problem with it. We use it for DB backups. It have 17 Gb and only we use 4 or 5 Gb for backups
- Rz9 have the orange light "on".

My questions?
- "Rz9 is used by informix in raw mode"
What mean that? Is critical?
- Can I move the data from Rz9 to Quantum disk without problems?
- Can I clone Rz9 disk in new disk?

thanks for all

[Vladimir Fabecic]

This disk is used by Informix DB and is not totaly dead (sys_check could read its disklabel).
And if Informix DB works fine, that proved it.
But it is time to do a full backup of your machine.
You will have to replace the failed disk anyway, soon or later.
Do "export" of your database.Then backup file systems using vdump, and backup raw partitions of rz9 using dd.
It would be best to do it in single user mode.
Then replace disk (rz9), label it with the same partitions and restore raw data using dd.
Do you know how to do it?
And do not worry much about Quantum disk. As I could see, it is working OK. Long time ago I was using Quantum disks on TRU64 and it worked OK.

[Vladimir Fabecic]

Sorry Marcelo, I replied before I read your last comment.
"Rz9 is used by informix in raw mode"
That means that database is using its own method for controlling device, not OS method (not file system). Is it critical? A bit more than in file system method.
You can not move the data from rz9 to Quantum disk just like that, you need help from DBA. I do not know much about Informix, but I think it is not big difference compared to Oracle DB.
You can backup that disk by using dd to tape, and restore to new disk using dd.

[Vladimir Fabecic]

I can suggest how I would fix the problem.
1. Get disk of same type as your failed disk (same p/n)
2. Do "export" of database
3. Plug new disk in next (now empty) slot in StorageWork drive shelf (SCSI ID 3)
4. Make device files for new disk
# cd /dev
# ./MAKEDEV rz11*
5. Go to single user mode (not too important, just to be sure no one is using device)
# init s
6. In single user mode do
# mount -a
7. Duplicate disk using dd
# dd if=/dev/rrz9c of=/dev/rrz11c bs=1024k
8. Shutdown system
# shutdown -h now
9. Swap physicaly these two disks in disk shelf
10. Boot system
>>> boot

I hope I did not forget something

[Michael Schulte zur Sur]

Hi Vladimir,

I would shutdown the machine before inserting the new disk.
You don't need a * for the MAKEDEV
And mybe you also need a
scu scan edt
before you can create the new devices.

greetings,

Michael

[Marcelo Farias]

Vladimir
Thanks for your help
Now... If I can´t obtain the same model disk like rz9...
How I do to?