)On 9/1/11 3:36 AM, David C. Rankin wrote:

---------------- Computer OS Disk (failed) 3Ware - 2 port striped (OK) ]______ mirrored mounted as /home/Data 3Ware - 2 port striped (OK) ] ----------------

Hi David and Joaquin, Actually, the config is this (my previous description may be a little confusing): ---------------- Computer OS Disk (failed) AMCC 9650SE-2LP Raid Controller (SATA II, PCI-e) 3Ware - 2 Disks mirrored (OK) ]______|-- Linux striped the 2 mirrors AMCC 9650SE-2LP Raid Controller (SATA II, PCI-e) 3Ware - 2 Disks mirrored (OK) ]______| (mounted as /home/Data) ---------------- Naturally, the controllers can only do Raid 1 or 0. I mirror 2 HD's on each controller. If one disk in the 2 mirrors fail, I can rebuild it with the HW raid controller. If 2 on the same controller mirror fail, I have a problem. The chance of this double failure on the same controller is slight. I have not yet in the last 6-7 years have a 3Ware controller failing. Not sure what will happen when one controller fails. I assume I can just replace the controller and it will pick up the config from the disks. Anyone know if a controller replacement will pick up the mirror? The 4 port 3Ware controllers (9650SE-4LPMLRaid Controller) do work if replaced, because the raid config is also on the disks, as far as I can hopefully recall correctly. I have not had a 3Ware fail yet, thus I lack the experience. I am considering replacing the 2x2 Ports with a 1x4 Port to have a HW-raid 10 without needing a Linux soft raid, for more safety; I just have to scratch the money together first. I will try with the openSUSE Live CD as Joaquin suggests, to make another backup of the raid (to be sure), and do a new install of the OS as David suggests. Thanks for the tips. :-) Al -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Thu, Sep 1, 2011 at 12:29, John Andersen wrote:

software raid for the ease of ... disk replacement.

How is that? Just the other day I replaced a failed SCSI drive in an HP server with the onboard RAID controller. You remove the failed drive while the system is running, and that is it. If it was a Linux software RAID I would have to run commands and hope they were correct. -- Med Vennlig Hilsen, A. Helge Joakimsen -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Brian K. White wrote:

Planning is wise, but that doesn't change the fact that it's an inherently inflexible solution that's waiting for you to slip up to burn you.

If you didn't happen to have any spare scsi disks on hand, most people today would not be able to lay their hands on one in less than 24 to 48 hours no matter how much they're willing to pay for overnight,

If ordered before 17:00 and in stock, delivery usually follows the next day with regular Priority Mail. (that's from a normal web-shop available to everyone). Regardless, a hardware array will typically be configured with 1 spare drive sitting idle, so you'd be replacing the failed drive a while after your array had already recovered. No need to hurry at all. -- Per Jessen, Zürich (18.0°C) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Fri, Sep 2, 2011 at 16:16, John Andersen wrote:

On 9/1/2011 11:38 PM, Per Jessen wrote:

...

Brian K. White wrote:>>> If you didn't happen to have any spare scsi disks on hand, most people

...

today would not be able to lay their hands on one in less than 24 to 48 hours no matter how much they're willing to pay for overnight,

If ordered before 17:00 and in stock, delivery usually follows the next day with regular Priority Mail. (that's from a normal web-shop available to everyone).

So posts the man from Zürich, a country of 15,940 square miles ;-)

...

Regardless, a hardware array will typically be configured with 1 spare drive sitting idle, so you'd be replacing the failed drive a while after your array had already recovered. No need to hurry at all.

Plus one on the hot spare(s).

Since they seldom looked at the server, and never rebooted it they got no warning of this at all.

If my Software Raid so much as hiccups, I get an email from mdadm.

If you buy all your disks for the array at the same time you have to be prepared for them to fail very close to each other.

FWIW: In my scenario it's a 1U server with only 2 drive bays. It took a week to get the replacement drive, but when the drive failed I got an email alert from Nagios, a free/open-source software to monitor servers. It's bad practice to setup a server and not put these measures in place. If you really value your data you shouldn't have some system in place to monitor it, be it you login every day and do the checks manually, or you implement something to do it for you on an automated basis. If you can't afford downtime, more important than what RAID you use, is how have you planned for when (not if) downtime happens? Wouldn't you feel silly if you setup the best RAID in the world and suffered a data loss because someone tripped over the extension cord between the server and the battery backup? -- Med Vennlig Hilsen, A. Helge Joakimsen -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On 9/2/2011 4:34 PM, Joaquin Sosa wrote:

If you can't afford downtime, more important than what RAID you use, is how have you planned for when (not if) downtime happens? Wouldn't you feel silly if you setup the best RAID in the world and suffered a data loss because someone tripped over the extension cord between the server and the battery backup?

Man you got that right! -- bkw -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

John Andersen wrote:

Since they seldom looked at the server, and never rebooted it they got no warning of this at all.

I take it they weren't using 3ware then? It happily sends emails on status changes and tells me when it's running a verify etc etc.

If my Software Raid so much as hiccups, I get an email from mdadm.

I use both 3ware and mdadm and am happy with both. I find 3ware a bit easier to drive but mdadm is more flexible.

If you buy all your disks for the array at the same time you have to be prepared for them to fail very close to each other.

One way to ameliorate this is to buy disks from more than one manufacturer, but it certainly is best to buy them at intervals. Cheers, Dave -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On 9/1/2011 12:29 PM, John Andersen wrote:

On 9/1/2011 6:49 AM, LLLActive@GMX.Net wrote:

...

I am considering replacing the 2x2 Ports with a 1x4 Port to have a HW-raid 10 without needing a Linux soft raid, for more safety;

In my humble opinion software raid is much safer than hardware raid if you are worried about controller failure or disk failure.

With Software raid: 1) Any compatible controller will do. On the Motherboard, on an add-in card, in a new machine. 2) You can mix drive types, even drive sizes. 3) You can span controllers. (your current setup kills the raid if one controller dies) 4) Replacement controllers don't get harder and harder to find as time goes on

Also, but not related to failure issues 4) Software raid improves as you upgrade the server - its not stuck at the speed you bought it 5) imposes very little load on the server (contrary to the FUD of the past) 6) easily configured in modern Distros, especially when your machine has 4 or 6 SATA ports. Yast does it.

I've been the 3Ware route. Its nice in that if done correctly your computer OS need never be aware that a raid is involved at all. But over the years I have migrated to software raid for the ease of maintenance, migration, disk replacement. I've used raid controllers with their raid software turned off (via jumpers) just for the ports in the past, but my current servers have scads of Sata/eSATA ports making this unnecessary. I still endeavor to put the drives of an array on different controllers although I've never seen any measurable difference in doing so.

Non of this is helpful in your current predicament, just a philosophical discussion.

+1 all the way. Used many hardware raid cards over the years on several varieties of unix. It was ok back in the old days when there was no other choice, but since years ago I can't stand the hassle of dealing with hardware raid cards. When you only have one machine, and only have one card, and it's nowhere near time to migrate or upgrade, a hardware card looks enticing. It's easy, and moving day is years away and you can afford the time to do a slow full copy to some different new hardware since it's just your one machine and you have nothing else to do and only yourself not a boatload of angry paying customers to worry about. If you have the luxury of living in a well funded or at least large, and utterly homogeneous, environment "We're an [HP|Dell|Intel|FooXYZ] shop, we have an account with FooXYZ and even though they don't make this FooXYZ card any more and even though it's all proprietary and only fits in this one FooXYZ case, we have piles of'em lying around..." then a hardware raid card looks enticing. If you're trying to grant the benefits of raid to an OS that can't do it in software, then a hardware raid card looks enticing. Any other situation and they're great right up until there's a problem and then they're just the reason you're telling the customer or boss any of a dozen different excuses instead of handling the problem without even resorting to backups. Perhaps after some googling and some experimenting with small ramdisk images just to be sure before doing something with the real disks. And/or after taking dd_rescue clones of the real disks instead of even risking touching them at all. Excuses like "I can't get one of those cards sooner than 4 days from now (by the way it's $800)" or "The card and drives are ok but I don't have another motherboard handy that has the required PCI-X slot for the raid card." or "The raid card firmware doesn't provide an option to do what I know needs to to happen to save 99% or 100% of the data and accept a tiny bit of corruption on a part of the disk that isn't even used by any files you wanted..." or "The boot message is ambiguous, and the cards manual doesn't even mention the topic, so we don't dare answer yes or no until we get the manufacturers tech support on the phone, and we can't reach them right now for some reason" or "raid6 would have saved the day but they didn't have raid6 when this card was made" etc... All of those show stoppers are not show stoppers if the array were software raid. Since it's hardware agnostic, you are free to piece together whatever solution is physically possible with any/all of the materials at hand (ie, the local BestBuy out in the middle of nowhere, or even less than that, just the other pc's in someones office and your and their homes in the middle of the night in the middle of nowhere) It's very difficult to encounter a situation that can't bet met somehow when every scrap of computer tech within reach could potentially be used. You have to know a little more but the important point is it's possible vs not-possible. -- bkw -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On 9/1/11 4:21 AM, Joaquin Sosa wrote:

Boot openSUSE 11.4 GNOME live CD and disk utility should see the RAID array and have the ability to start the RAID and mount the resulting volume.

Why such the odd arrangement? It would be less headache to use RAID 0+1 native to your RAID card. It is a 2-Port card, i.e. only Raid 0 or 1. The issue is your hardware RAID cards are worthless in your configuration. Say you used desktop-grade hardware in your system and it develops bad RAM, it the MD-RAID driver will write the corrupt data to the array. ... ok, but how will bad ram not do the same to any other set-up?

I was a bit nervous about the SW-Raid. With HW-Raid, you open the 3Ware Bios, rebuild and start the OS. Painless. I have not used SW-Raids before. Today I succeeded in recovering the SW-Raid stripe. In YaST -> Services -> Partitioner I just redefined the two HW-Mirrors as a Raid 0, without any formatting or mount-point. When I first looked in the FS with Gnome and KDE Data Managers, there was nothing on the /home/Data volume when I mounted it. On Gnome Nautilus there was a new entry on the left of an unmounted Volume, named exactly as I named it in the previous installation; I felt a heartbeat skip. It wanted some authorisation before it could mount it. It failed, because I used the original root password, which it did not recognise. I went to the CLI Terminal and su -, then looked, and it was all there, I made another full backup to a USB HD, just in case. Now I can try some alternatives. Phew.... I was relieved. It seems the SW-Raid also has the raid info on the disks. Al -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org