Bootable Backups? Possible? Why not?

[commissar_mo]

Objective: My OS drive I want to be able, however imediately boot a drive which is identical to the dead drive, and has been 'slaved' to it for its duration (i.e. it has been incrementally backed up to), and that's it.

---

So in essence, my main issue here is the creation of bootable backups. I've spoken to mass consumer software company REBIT support, as well as read over different backup software forums, and it seems to me there is an understanding of this out there, but it's a bit hazy (judging by the wide range of opinions I've heard on it)...

1. People talk of some software copying the MBR while others don't. I'm not sure though if copying the MBR of a drive is all that's required to make it bootable. (Likely, I don't understand what truly makes an OS 'bootable')

2. If indeed, it is, then what I want is possible. I have an external drive running 'cloning' software (which is NOT an image, I understand, as an image is by definition a snapshot at a single instant in time). When my drive dies... I power off, swap my clone drive and boot up... the hardware AND software is none the wiser as to what just happened.... everything is back to normal.

It seems really simple, but I know there's something I'm missing here... something about the original drive, be it the MBR, meta data, etc.... makes it distinctly different from the 'cloned' drive... and I just can't figure out what...

It's bothering me because, IF the drives were, in theory, to be exactly (EXACTLY) the same, as the term 'clone' implies to me, the hardware/software should have absolutely no idea which drive is which... I should be able to freely interchange them (and boot, etc.) without even noticing...

I'm really interested in finding out IF that's possible, or perhaps more relevantly, if NOT - Why?

 

[WyomingKnott]

A long question deserves a long answer. (I tend to ramble)

It is indeed possible. There are probably somewhere between three and eleventy-seven ways to accomplish this. Plus one that you didn't mention...

Two questions are "How many (minutes/hours/days/weeks) of work and OS configuration are you willing to lose in an 'event'?" The other is "How much time is acceptable for the replacement drive to come on-line after the system drive fails?"

If the answer to both is "None," then this is a really easy question to answer. Do a fresh OS install on a RAID1 pair. I will not describe how or the intracacies here. Suffice it to say that RAID1 means that every write is done to two identical drives "at the same time." If one fails, the other is still full of all your OS, files, and data, and your system keeps running with no interruption. With a good setup, an alarm will go off warning you to replace the failed drive.

If the answer to the first is "A day or two," and the answer to the second is "At least a half hour," then you can "clone" your drive at the end of each day to a backup drive. Pull the dead drive, put in the backup, and go. Yes, there are Master Boot Records (one per drive), Partition Boot Records (one per bootable partition), boot loaders, active flags, and other stuff that don't get copied if you do a normal copy from one drive to another. But there is endless software out there that can do a competent clone of an entire drive. EASEUS gives it away for free in the hope that you will buy one of their licensed products. I'm still using the Norton Ghost version from 2003.

It's only if the answer is somewhere in-between that the question becomes complicated. Since it seems clear to me that you want zero loss of OS and data, then read up on RAID and then set up RAID1 for your boot drive. My personal approach is different. My OS lives on one drive, and is backup up infrequently. My data lives on another drive, and is backed up often, but not as oftern as it should be. Certain key directories on my data drive are mirrored in realtime to a partition on my system drive (if both fail, I am hosed, but I decided to work towards protecting from a single failure) with MirrorFolder.

If my OS goes down, I put in a new bare drive and restore my last OS image backup. Half-hour, tops.

One other thought - MirrorFolder might work for you, allowing you to mirror your OS drive. But the overhead is greater than running RAID1. If you do run RAID1, remember that it is no substitute for frequent backups. It won't protect you against a zillion other vulnerabilities, such as malware and deleting stuff it turns out that you needed and the mysterious disappearing folder monster.

Edit - OMG, what an idiot! I wrote RAID0 (more vulnerable) instead of RAID1 (safer). You would be very unhappy with RAID0 - either of the two drives fails, you lose the data on both.

 

[sminlal]

If you choose a RAID 1 solution, be sure to test it before you commit live data to it. I've worked with RAID controllers in which it was impossible to take a split former RAID set member and put it back into set with a second mirrored drive without initializing both drives. A controller with that restriction would let you perform ONE restore, but after that you wouldn't be able to re-mirror it.

 

[commissar_mo]

Thanks for the explanation!


*I do indeed currently use a RAID 1 for a media (warehouse drive I call it) disk. One of the issues was I don't like the write performance of RAID 1, so I put my OS drive on RAID 0 2 years ago. I had everything backed up... but didn't really expect the MOBO controller death to annoy me as much as it did. I've always used RAID 10 for my data on a hardware controller, but have been concerned as of late of controller failure, and wanted essentially extra security (bootable clone disk) to not even have to worry about replacing the controller.

**I essentially want to AVOID RAID now because while it has disk redundancy (sans 0) it still has controller-death. I have been informed that swapping identical hardware controllers is fairly robust, so while I'm not as concerned about it, I'm still looking for paranoia-salves.

That said... After my RAID 0 crash, and 3 days on these forums, I think this is my plan:

1.

A) I think the potentially best balance between performance, redundancy, and up-time (ignoring cost) would be to run a RAID 10 system for my OS AND keep my RAID 10 for my data. I hear the delta between RAID 10 and 0 performance isn't really statistically significant over a heavy conglomerated system use (e.g. I play games, edit films, run productivity software, Maya and some rendering, and the usual gamut of basic use), though I imagine the write performance suffers, even if trivially, from the dual-disk write requirement...

B) Originally, as I have in the past, I was going to run the OS on RAID 0, clone it, and not care if the RAID 0 died, but benefit from the increased performance while it lived; now though I'm more concerned with up-time, so I'd rather take the perf hit (if there is a noticeable one) and have my OS drive shrug off a disk failure.


2. (Mainly just for fun, I'm not that insane) I'd like to see if you can run a hot spare with RAID 10 - that way, even a disk crash in the RAID 10 would automatically heal itself and keep chugging along.


3. On top of these two RAID systems, I would then as you suggest, use "Imaging/Cloning/Backup" software of some kind (that does copy the MBR and provides for a bootable clone/image disk) to make an image of both RAID systems, say weekly or whenever I install/add/delete apps or settings.

4. I would simultaneously run the incremental backup feature of said software (I think Acronis has this for instance) to catch (mainly data) changes between images.

-------------------------------------------------------------------------------
This provides both uptime, performance, and data redundancy.

The (relatively speedy) striped OS RAID10 system can go down two mechanical ways - disk failure (which the controller will signal and rebuild with a hot spare or a swap) which won't interrupt service, or a Controller failure, which will.

The controller failure can be dealt with by swapping the controller card. If, as sminlal below has mentioned, there are controller card mishaps in rebuilding the array, which I'll fearfully assume will happen (in this case RAID 10), the image can be brought to bear, restored and then updated with the file changes saved by the backup software.

That is, so far as my research allows, the most bulletproof home data system I can think of (assuming NAS, Online, and external offsite backups for various other stuff).

Many thanks.

 

[WyomingKnott]

That's a _home_ system?

Finally, someone who takes reliability and backups even more seriously than I do. Nice to see you.