Mail Server Woes

BenH

I've just been asked to go and look at a Suse 3.1 mailserver whose SCSI drive is currently in the process of dying after spending the past year buried under a pile of hubs, in a warm, draughty and humid room.

I've been told that theres a second, unused SCSI drive in the machine, so I had the thought of copying the entire root directory straight across and then edit the boot loader to spin down the first scsi drive and boot from the second and hey presto the servers saved with only a single reboot.

Normally I'd dd the disk to a network mount and install a new disk before dding it back, however I dont thing that will be an option.

So is my plan plausable, or nothing more than deluded ravings of an overworked, underpaid madman?

Cheers,

Ben

SkepticOne

It is worth a try. Is the working drive partitioned? If so, bear in mind that fstab will need to be updated to reflect the new drive.

I would try to get a replacement drive, though, as the one in the machine may have suffered the same deterioration as the working one.

Be careful with dd and bad blocks.

BenH

Originally posted by SkepticOne
It is worth a try. Is the working drive partitioned? If so, bear in mind that fstab will need to be updated to reflect the new drive.

To be honest I have no info on the second drive, but it should be formatted as reiser, so its just a question of firing up the partitioner

I would try to get a replacement drive, though, as the one in the machine may have suffered the same deterioration as the working one.

I know the guy who did the install so it should have been put to sleep if not in use and I dont think they have the money for a replacement.

Be careful with dd and bad blocks.

Yes, if the other disk is deteriorated then I'll just back up the mail spools and user accounts.

Thanks.

Typedef

Copying recursively the entire / might be impracticle, on the /dev directory.

Also the bootloader etc, won't take well to that (it'll need a reinstall most likely).

Just make sure you copy symlinks as symlinks, as opposed to the files they point to.

SkepticOne

A quick and dirty approach would be:

Activate the second unused disk, then:

dd if=/dev/sda of=/dev/sdb

Then change the scsi ordering of the disks so that the new disk assumes the device name of the first.

This assumes that a) there are no bad block problems on either of the disks b) they are named sda and sdb respectively, c) they have the same geometry.

If the boot loader resides on sda then it should be copied over to sdb and still work. If it resides on a separate disk then it should also still work as it will still be referring correctly to sda.

Personally I prefer to manually create the partitions and file systems and copy over files, reinstalling the bootloader rather than transfer the disk image with dd.

edit: back up everything, obviously.

SkepticOne

Appologies,

dd should only be used on unmounted drives. Obviously if sda is the the current boot drive then you can't unmount it.

BenH

Thanks for your responses. Turns out that the second disk has been partitioned and has been spinning its merry way around at 10KRPM doing nothing but generating heat.

To be honest I think I'll export the users to squirrelmail, backup the mboxes, possibly use knoppix to dd one disk to the other, or rsync it to one of my network drives and then give the server last rites.

Were not getting paid for this so they should be grateful (ha) for whatever they get.

Thanks again,

Ben

BenH

Just a little update,

The server room varied in temperature from 25 - 38 Degrees C, spiking at 45. The floors were covered with masonry dust, and there were three entrances into the room, the doors of which were permanently open.

Their now going to claim on the warranty of the drives and I wish there was an emoticon for repeatedly banging my head against a desk

Typedef

I had dodgy hardware like that once.

Dropped the fuppin server off a cliff ... and a 50 foot drop...

Scratched the paint....

Demanded a supercluster as a replacment for DOA poweredge.

BenH

Originally posted by Typedef


....

Demanded a supercluster as a replacment for DOA poweredge.

At the very least!

A friend recently asked me to look at his scanner, he'd moved the computer recently and said beofre it had worked brilliantly, after the move it didnt work at all...

After 20 mins of me trying to get the stepper motor working, he finally confessed to having dropped it twice.

dazberry

Originally posted by BenH
Just a little update,
The server room varied in temperature from 25 - 38 Degrees C, spiking at 45. The floors were covered with masonry dust...

[OT]Actually that would be an interesting thread in itself - what's the worst server room you've seen? Not OS related, just the holes they want to put their servers in 'cos health and safety wouldn't let them put people in there instead

D.