#296777 - 26/08/2007 18:06
Re: Hardware RAID Advice
[Re: Seth]
|
carpal tunnel
Registered: 29/08/2000
Posts: 14492
Loc: Canada
|
Quote:
The article had three drives in the RAID1 array. Since RAID1 is mirroring, I've never seen 3 drives in a mirrored array. So I was wondering if one drive is actually mirrored to the other 2 in the array.
RAID1 arrays of four drives are actually quite common in the commercial systems I work on here. These people are really *paranoid* about their data, and don't want it ever to be exposed with no hot backup online. With only 2-drive RAID1, when one drive fails they would then be without a live backup until a new drive got "built" back into the RAID1. With three drives, no such problem. With four drives, even safer.
Quote:
So it sounds like this...someone correct me if I am wrong:
- All writes to the boot partition are mirrored to two other drives.
- If any one of the three drives fails, the fourth will automatically join the array so that the first partition is still mirrored to two others.
- If the first drive fails, then the system would boot off drive two which would then be mirrored to drive three and four.
Yup, that's the idea. Many modern BIOSs will automatically do that last step (boot from first available drive) in particular.
-ml
|
|
Top
|
|
|
|
|
#296778 - 26/08/2007 20:40
Re: Hardware RAID Advice
[Re: mlord]
|
carpal tunnel
Registered: 10/06/1999
Posts: 5916
Loc: Wivenhoe, Essex, UK
|
Quote:
Yup, that's the idea. Many modern BIOSs will automatically do that last step (boot from first available drive) in particular.
Does anyone have any idea on how I can actually test that step in my Linux software RAID setup. I have tested disconnecting the primary drive, but that doesn't seem to be testing quite the same case of the actual failure of one drives to spin up. I can't work out how to simulate that case though.
_________________________
Remind me to change my signature to something more interesting someday
|
|
Top
|
|
|
|
|
#296779 - 26/08/2007 21:24
Re: Hardware RAID Advice
[Re: andy]
|
new poster
Registered: 15/04/2007
Posts: 15
|
How about just disconnect the power. I know, that isn't quite the same thing, but I've struggled with that same question myself--how to test real life failure scenarios.
|
|
Top
|
|
|
|
|
#296780 - 26/08/2007 22:25
Re: Hardware RAID Advice
[Re: Seth]
|
carpal tunnel
Registered: 10/06/1999
Posts: 5916
Loc: Wivenhoe, Essex, UK
|
Disconnecting the power to the drive isn't the same thing at all, it is no different to disconnecting the SATA cable, which I have tested.
_________________________
Remind me to change my signature to something more interesting someday
|
|
Top
|
|
|
|
|
#296781 - 27/08/2007 00:13
Re: Hardware RAID Advice
[Re: andy]
|
carpal tunnel
Registered: 29/08/2000
Posts: 14492
Loc: Canada
|
Quote:
Disconnecting the power to the drive isn't the same thing at all, it is no different to disconnecting the SATA cable, which I have tested.
There's no good way to test every possibility. Fortunately, with most drive (it seems), failure to spin-up means the drive will also fail to complete the "IDENTIFY" handshake, and a good BIOS will then assume the drive is not present and try the next.
But there's always a possible pissy spot there, and without a failsafe BIOS (source code generally required to verify that..), one never knows..
Cheers
|
|
Top
|
|
|
|
|
#296782 - 31/08/2007 04:36
Re: Hardware RAID Advice
[Re: andy]
|
new poster
Registered: 15/04/2007
Posts: 15
|
Quote:
Disconnecting the power to the drive isn't the same thing at all
That is why I said it wasn't really the same thing, but how else are you going to test that without hacking a BIOS to control how a drive initializes at boot?
|
|
Top
|
|
|
|
|
#296783 - 31/08/2007 06:23
Re: Hardware RAID Advice
[Re: Seth]
|
carpal tunnel
Registered: 10/06/1999
Posts: 5916
Loc: Wivenhoe, Essex, UK
|
Quote:
but how else are you going to test that without hacking a BIOS to control how a drive initializes at boot?
I don't know, that was the whole point of asking if it was possible and if so how !
_________________________
Remind me to change my signature to something more interesting someday
|
|
Top
|
|
|
|
|
#296784 - 31/08/2007 10:06
Re: Hardware RAID Advice
[Re: andy]
|
carpal tunnel
Registered: 24/12/2001
Posts: 5528
|
Quote:
Quote:
but how else are you going to test that without hacking a BIOS to control how a drive initializes at boot?
I don't know, that was the whole point of asking if it was possible and if so how !
I'd say a rubber mallet would work...
|
|
Top
|
|
|
|
|
#296785 - 31/08/2007 12:17
Re: Hardware RAID Advice
[Re: tman]
|
carpal tunnel
Registered: 18/01/2000
Posts: 5683
Loc: London, UK
|
Quote:
I'd say a rubber mallet would work...
Angle-grinder? Through the platters?
_________________________
-- roger
|
|
Top
|
|
|
|
|
#296786 - 26/10/2007 02:52
Re: Hardware RAID Advice
[Re: andy]
|
new poster
Registered: 15/04/2007
Posts: 15
|
Quote:
I don't know, that was the whole point of asking if it was possible and if so how !
I had some unrelated issues on another system here at home which brought this thread back to mind. You were asking about different boot scenarios and how the BIOS will handle them. Without going into the details of my issues here, let me say that I think there is no way to deal with every possible contingency to ensure the system boots properly.
Here are a couple situations I am quite confident will prevent your system from booting despite all the efforts to make everything redundant:
1) Drive spins up and identifies itself to the BIOS but is not functional, or at least not really functional enough to boot from
2) Drive is the master (or is on a SCSI bus) and is malfunctioning to the point of making the channel/bus unuseable. I had this happen where one drive (the second on the bus) went bonkers (not even a real crash), and prevented the other two drives from [effectively] communicating with the controller. And when this happened, half of the boot attempts hung without even giving a "non-system disk" error.
3) Any other type of failure that allows the BIOS to see the presence of the disk, but disk is not accessible.
All of those will likely give you one of those "non-system disk or disk error" messages on boot up. The best failure scenario is where the drive is completely dead or non readable by the BIOS, in which case it most likely will move onto the next device. As I understand it, since drives became auto-detectable by the BIOS years ago, this mechanism actually works by storing the parameters in a sector on the disk, and the BIOS reads that. So if the drive really fails to spin up, the BIOS can't possibly identify it.
It occurred to me that one possible way of testing how your BIOS reacts to this type of scenario where the device exists, has power, but is "not spinning up" is to test something similar but different. If you had two drives on different IDE channels, you could deliberately misconfigure the first to be a slave instead of master on its channel, or something of that nature. If that doesn't work, maybe there is something else along those lines you could do to essentially accomplish the same thing. Then you would see what the BIOS does in that circumstance, i.e. move onto the next drive or just choke.
Just some thoughts I wanted to pass along.
BTW, for those interested, I continue to be more and more a fan of this software RAID approach using a generic Linux distro. I was enamored first by the hardware RAID, then by the specialty NAS OS (with software RAID) approaches, for different reasons. I thought it nice to have a good customized NAS GUI with slick performance graphs and so forth, but in the end none of those options stacked up to just good ol Linux. Far more robust and resilient than anything else I have experimented with. Alas, I have to do some of the customization work and configuration, but in the end I have what I need and it works like I want it to.
|
|
Top
|
|
|
|
|
|
Previous Topic
Index