LVM Problems

[TwoProc]

On this D-class server, I think you've lost a major chunk of hardware. This is just what I see from here without looking at it. To tell you truth without some sort of confirmation (from HP or someone who is better at hardware) I think it may really bad.

Looks like a whole piece of the I/O sub-system is at least partially hosed. You can play around, but you're going to risk losing data.

The safest thing is to shutdown and move all of the disks over to another server (like the other one in the drawing), and see if everything comes up. If it does (and I think it will), get a backup of everything and then ask management "which server they want up" if you don't have a spare or test server - because I don't think you're going to get both. But, HP (or somebody who's better than myself at hardware) can tell you that better than me.

In any case, I'd start using another server I could spare to start troubleshooting, by putting all of the disks from this server into another almost exactly like it and see if it will come up.

If you've got no spare test servers, then I'm stuck as to tell you to go next, other than calling HP and getting a billing for time and materials. But I think you've got a "D" and you need to decide if beginning troubleshooting at $300 to $400 an hour so (plus travel) plus parts is going to be worth that little server.

You can buy them used cheaply, but I'm not sure about in your area...

John

[Leila Maria Rebel]

John, isn't it similar to start the services in the adoptive node?

[Leila Maria Rebel]

The command "dd" works fine for c1t1d0 and
results an error for c0t1d0.

[Leila Maria Rebel]

Chan, are you still there?

I'd like to know how to use insf without recreating the special file.

I've foun a thread in this forum similar to my problem and the solution was

insf -e -C disk.

Could you help, please?
Regards, Leila.

[Adisuria Wangsadinata_1]

Hi Leila,

The problem looks like due to the faulty disk (c0t1d0) and you need to replace this harddisk, because the dd command result already gives you the I/O result.

I believe there's nothing wrong with your connection and the SCSI card, since the situation that made by these will be worse (cause will happen on the affected bus, not just a single disk).

Now is how to makes the VG03 online even with single harddisk (c1t1d0) and later put the replacement for the mirror harddisk (c0t1d0).

You can try to break the mirror first and see if you can activate + mount the file system into the OS on the single disk.

Meanwhile prepare your backup for vg03.

The simple method is to recreate the vg03 and restore the data from the backup (if the valid & uptodate backup is available).

Hope this information can help you.

Cheers,
AW

[Devender Khatana]

Hi,

The data should be intact atleast in one of the drives (One mirror copy). A close look at various parameters show that the devices attached to vg03 are /dev/dsk/c0t2d0 and /dev/dsk/c1t2d0 whereas the error is coming in other two disks i.e. /dev/dsk/c0t1d0 & /dev/dsk/c1t2d0.

The error reported is usual if the two disks are being used for some other purpose than LVM. If the two disks belongs to LVM then I would suspect that two disks has been replaced physically without notifying the changes to the system.

Also out of 4 volumes configured into vg03 only 2 are available making only one LVOL available (LVOL5) you can try mounting it and everything in this LVOL should be fine.

You can check from "strings /etc/lvmtab" for two other LVOLs configured in this VG and try to see any error with them. If the disks are not yet physically replaced I am sure you will be able to recover data still.

Post
#strings /etc/lvmtab


HTH,
Devender

[Leila Maria Rebel]

AW, good morning!

Could you help me on the steps for breaking mirror and setting up VG03 again?

Regards,
Leila

[Devender Khatana]

Hi,

Good Evening as it is evening here in India.

Breaking and recreating mirrors will not help as two disks are missing and they hold both copies of four LVOLs.

Post #strings /etc/lvmtab

as mentioned earlier.

HTH,
Devender

[Carl Munnelly]

Just confirm is this a D Class and what O/S is running?

If a D class has O/S been recently updated to 11.
HP-UX 11 does not support EISA h/w and will therefore discable it, meaning half system port unavailable

[Leila Maria Rebel]

Hello, Devenda!

Follow the lvmtab.

[Leila Maria Rebel]

Carl, thank you for your message!

The OS is HP-UX 10.20.

Leila

[Devender Khatana]

Hi,

Luckily the two disks found extra are the same two which are missing from vg03. You will have to try vgexport/vgimport to try recovering the whole data fro either copy prior to replacing the disk.

Note down minor number of the VG by "ll /dev/vg03/group". Also post the output of
#pvdisplay -v /dev/dsk/c0t1d0
&
#pvdisplay -v /dev/dsk/c1t1d0


#vgexport /dev/vg03
#mkdir /dev/vg03
#mknod /dev/vg03/group c 64 0x??0000
Where 0x??0000 is the minor number noted from above.
#vgimport /dev/vg03 /dev/dsk/c0t1d0 /dev/dsk/c0t2d0 /dev/dsk/c1t1d0 /dev/dsk/c1t2d0
#vgchange -a y /dev/vg03
#vgdisplay -v /dev/vg03

HTH,
Devender

[Leila Maria Rebel]

Devender, the pvdisplay commands result errors.

Follow the output.

[Devender Khatana]

The error is understood, this is because these two disks are not attached to any VG. The option left with you now is to try vgexport/vgimport as mentioned earlier.

Also do a
#vgchange -a n /dev/vg03 before doing vgexport.
HTH,
Devender

[Leila Maria Rebel]

Devender,

I've tried vgimport before, but I've got "PV does not contain any LVM information"

By the way, I'll try again.

Is there any risk of losing data?

Regards,

Leila

[Devender Khatana]

Hi,

The data will not be lost by doing so. But the current stage shows that some data is allready lost and is not accessible. If the vgimport reports that the disk do not contain any LVM information I would suspect that the disks has been either physically replaced or the information is lost due to drive problems.

lvol5 (Which is off more than 4GB) should be accessible at present and you should back it up before doing something else. Also I will be leaving for the day in another 45 minutes.

HTH,
Devender

[Leila Maria Rebel]

Hello, all!

Unfortunatelly, I had to give up.
My customer established a deadline and my time is over.

I recreated the whole structure of the VG, excluding the damaged disk.

The last backup was recovered.

I'd like to thank you all for helping me!!!

I hope there won't be a next time ;-)

Best regards,
Leila

[TwoProc]

OH well, sorry you didn't make it. Thanks for the points, and better luck next time.