Strange RAID 1+0, 0+1 Question.

[CaptRobertApril]

Ok, bear with me as this is a bit confusing.

The system will have:

6 x 7200.10 750GB for data
1 x 160GB Raptor for OS

Now the config of the 6 data drives is the strange one.

The system will be set up in RAID 1+0 or 0+1. At any given time only four data HDs will be in the case.

HD 1 & 2: Actual Original Data
HD 3 & 4: Data Mirrors

So first of all I'd like to find out whether 1+0 or 0+1 is the best setup. I've read the wiki and I'm still confused.

Now the important question is this. Once a week,

HD 3 & 4

Will be swapped out with

HD 5 & 6

So that the way it would work is:

Week One

HD 1 & 2
HD 3 & 4

Week Two

HD 1 & 2
HD 5 & 6

Week Three

HD 1 & 2
HD 3 & 4

The two HDs that are out of the case are stored offsite for security.

Now.......... it seems to me that at the end of week one when I swap out HD 3 & 4 for HD 5 & 6, I'd have to overwrite all of the newly installed drives with the data from HD 1 & 2 otherwise, the data on the mirror drives will be all messed up. And I'd have to repeat this process each week when I swap out the data drives again.

Am I right? Wrong? Insane? All of the above?

 

[SomeJoe7777]

OK. My opinion is ... No. No. No. And, ummm, no.

RAID is not the same thing as backup.

RAID is a tool for preventing data loss due to hard drive failure, and to keep the server online while a drive has failed.

Backup is a tool for keeping past saved copies of data accessible for a period of time in order to facilitate going back to a previous version, should the current one become unreadable, corrupt, deleted, infected, etc.

Use the correct tool for the job. If you want an offsite copy of the data, you need a backup solution. (Tape is ideal because of the portability and cheap media cost/GB, external hard drives can be used as well).

Using RAID as you have outlined it is extrordinarily dangerous. Broken mirrors have to be rebuilt, which is a drive intensive operation that you would be doing once a week. If something goes wrong during the rebuild, you could conceivable lose all data.

For the differences between RAID 0+1 and RAID 10, see my previous posts on the subject. RAID 10 is superior to 0+1. However, depending on the RAID controller you have, RAID 5 might be more appropriate anyway.

 

[Dante_Jose_Cuervo]

Dude, if you really wanted to do something like that for backup you'd have to like use RAID 3 or 4... and swap out the Parity drive... but that's all I can think of. You see, RAID isn't backup like SomeJoe7777 said... it's just meant to keep you online if a drive fails... but it can only do so much... If I were you I'd get the above-mentioned RAIDs or RAID 5EE...

 

[CaptRobertApril]

Thanks for the input, SomeJoe and Jose Cuervo (BTW, I love your drink...)

I appreciate the clarification, especially when it comes to doing such a drive-intensive op once a week. It seems that there would be no option but to completely rewrite the drive(s) which is gonna be a royal pain in the a$$.

Prices for HDs have come down to the point where they're quite competitive with Tape (at least once you stay lower than the multiTB levels) so I'd rather stick with them.

The problem is that I require:

a) Constant real-time backup
and
b) Once a week offsite backup.

To do a), it would seem that RAID 1 is just fine. No problems there. I think the problems come into play when I want to throw RAID 0 into the mix and then back up the whole mess for b).

In that case, any suggestions should I want to abandon the RAID 0 altogether? Should make it considerably easier.

 

[Dante_Jose_Cuervo]

Well... if you wanted to get some good backup you could just get a RAID 5 or 6m, or dare I say it... RAID 5EE config and have at least 1 layer of redundancy. Now I think that you could do a RAID 10 and then swap out the drives... though you'd have to have it rebuild everytime you did that.

 

[CaptRobertApril]

Yeah, I see what you mean, basically using the hot spare drive as the replaceable on 5EE. But I'm still not clear as to what procedure I should use at the end of the week when I need to swap the offsite HD. Would I still have to rewrite the newly inserted drive from scratch?

 

[Dante_Jose_Cuervo]

Well, for something like you're doing I'd say to just go with like a RAID 10 config and just taking out one set of the mirror at the end of each week. Then you could replace those with blank drives and it would rebuild... if needed... you could use the old drives as reference... but I'm not sure how the config would handle that... I think that it would corrupt almost instantly if there were any errors found.

 

[CaptRobertApril]

I think that it would corrupt almost instantly if there were any errors found.

That's exactly what I want to avoid. I think i want what doesn't exist. RAID 5237! Hotswap mirroring with automatic incremental backup. :lol:

 

[gse1]

Well first of all forget about using RAID and swapping out your disks once a week. This is not what RAID was intended for and you are asking for trouble doing this.

The only time you should consider RAID 0 is for performance reasons. If you don't care about performance then forget about RAID 0. You could argue that you can use it two tie the filespace for two disks together but with 750Gb drives that doesn't seem to be an issue.

RAID 5 is a better consideration than RAID 1 in this case. You can configure a RAID 5 array with one parity disk and a hot spare which means 3 of the 6 disks would have to die before you lost the filesystem. However, you have to pay for a RAID 5 controller to do this and this may be pricey. Plus this doesn't give you off-site storage.

The best way of achieving what you want with you have already is to use RAID 1 to mirror disk 1 to disk 2 and disk 3 to disk 4, buy 2 external hard drive bays for the other 2 disks and run a backup program to copy the data onto these disks, then take them off site.

These are pretty specific requirements you have, can I assume this is for a work PC rather than a home one? What are you actually trying to achieve?

 

[CaptRobertApril]

The best way of achieving what you want with you have already is to use RAID 1 to mirror disk 1 to disk 2 and disk 3 to disk 4, buy 2 external hard drive bays for the other 2 disks and run a backup program to copy the data onto these disks, then take them off site.

These are pretty specific requirements you have, can I assume this is for a work PC rather than a home one? What are you actually trying to achieve?

Definitely agreed that the simpler way is the best in this case. It may be paleolithic, but just running a conventional backup onto the two offsite disks may be the way to go. It's a bit 1985 floppy-net, but if the option is to go with something pricy, complicated and set up for catastrophic failure, then off smiling into prehistory we'll go.

My home and work PC are one and the same. The number of apps I'll run in the average week can run up to 20, including 15 brand new ones. Everything from image editing to massive CAD files to ??? and lots and lots and lots of video, including some hidef. Therefore I want 1.5TB of fast main HD, have it backed up automatically in real time (RAID 1) and then have a once a week offsite backup of the whole shebang.

 

[gse1]

Well if performance is an issue than what you want is indeed RAID 0+1, two RAID 0 logical disks coupled together using RAID 1.

And just because this solution was used 20 years ago doesn't mean it's not a good idea to use it today. If it ain't broke don't fix it and all that.

One other possibility is if off-site backup is a paranoia step. You could have a seperate low-cost fileserver, stick disks 5 and 6 in that and have a job that automatically backs up files to that machine over a network. Physically seperate from your PC so you keep your data if your PC explodes, but not offsite so if your house explores you are still knackered.

 

[CaptRobertApril]

Well if performance is an issue than what you want is indeed RAID 0+1, two RAID 0 logical disks coupled together using RAID 1.

And just because this solution was used 20 years ago doesn't mean it's not a good idea to use it today. If it ain't broke don't fix it and all that.

One other possibility is if off-site backup is a paranoia step. You could have a seperate low-cost fileserver, stick disks 5 and 6 in that and have a job that automatically backs up files to that machine over a network. Physically seperate from your PC so you keep your data if your PC explodes, but not offsite so if your house explores you are still knackered.

In the neighbourhood where this PC lives a dude coming home from a bar last weekend was shot three times in the forehead for a few bucks and his Nike Airs. A $10K PC can be fenced for two grand and that will buy a good long fix for some horse addict. That's why the BU disks live ten miles away. If some bugeyed strungout freak breaks in, I want to hand him the PC, wish him the best of luck, show him the door and call my insurance agent without the added ballast of a few lead bullets in my brain.

The only ADSL we can get in this area is 4MB down 256K up so to upload 1.5TB should take about seventy-two thousand years, give or take a day. Unfortunately the physical "bring the HDs to the PC, back 'em up, take the HDs back out" scenario is the only one I can come up with.

The question between RAID 1+0 & 0+1 is what this thread started with. Looks like it still is the way to go. This way, I'll have 1.5TB of fast data, 1.5TB of realtime backup, and then when it's BU day, I can just put my feet up and watch a lot of pr0n while the HDs whirr away...

 

[Jim_L9]

If you can initially sync up the 2 remote hard drives and keep them on a server somewhere else you can use rsync http://www.gaztronics.net/rsync.php to keep the data up to date on the remote HD on a server.

 

[CaptRobertApril]

If you can initially sync up the 2 remote hard drives and keep them on a server somewhere else you can use rsync http://www.gaztronics.net/rsync.php to keep the data up to date on the remote HD on a server.

Problem is that there is no way to connect the PC with the server since the latter needs to be 10 miles away, and the only option is a very slow ADSL.

 

[ranger90]

Ok, bear with me as this is a bit confusing.

HD 1 & 2: Actual Original Data
HD 3 & 4: Data Mirrors

So first of all I'd like to find out whether 1+0 or 0+1 is the best setup. I've read the wiki and I'm still confused.

I understand what you're doing. You're swapping out a member of each mirror set and storing offsite.

This is do-able. A hassle, and has risks, but is do-able.

First you'll want to set it up that you "stripe the mirrors". Do NOT "Mirror the stripes". There is a difference. If you stripe the mirrors, you can loose a mirror member from each mirrorset and be OK. Which is what you want to take advantage of.

Second. You have to be very, very careful when you split your mirror sets for disk swap-out. I'd be willing to wager you will need to keep your disk order. Let me explain. Disk 1 and 3 are mirrored, becoming mirrorset "A". Disk 2 and 4 are mirrored, becomming mirrorset "B". Then you stripe Mirrorsets "A" and "B" for your final "Disk." Now, say disk 1 and 2 blow up. You have to revert. Make sure up plug disk 3 back in its original place, and disk 4 in it's original place. Then rebuild the mirrors back onto "new" disks 1 and 3. But also, this is to maintain the relationship of the "stripeset" between mirrorsets "A" and "B". The data on the mirrorsets are written in order. and if you accidently swap disks 2 and 4, the order of the data will be reversed and in all likelihood trashed.

Third. You must make sure of data quiescence before splitting mirrorsets. If there's data in disk caches, it must be written to the spindles before any split can occur. This often requires a "sync" in the Unix world. And/or a disk dismount in other operating systems.


Finally. Before you put this into production. Test, document, test, and document. You definitely will need to know if what you propose will work. If it's recoverable. With the above scenario, can you rebuild mirrorset and stripeset.

If your requirement is maximum uptime, I don't suggest this as a solution. Data integrity is paramount, and quiescing disks would probably disqualify this as an option.


An alternative solution would be to opt for external high speed storage. The raid article here today mentioned the RocketRaid. Use one controller, and two external storage boxes. On a weekly basis, copy internal data to external storage. Each week, alternate between external storage sets.


Hope this helps.

 

[ranger90]

The problem is that I require:

a) Constant real-time backup
and
b) Once a week offsite backup.

RAID really doesn't give you backup. Yes it can be used to get a backup, but RAID only really provides data security. A "backup" is an copy of the data kept out of the way from normal usage. To provide constant real-time backup. You'd need some sort of journaling file system that writes two separate and distinct files of the files in separate distinguishable areas. One file in the work area, and the second in a hot storage area. The storage area would not be read from except in the course of a restore. The new Mac OS is supposed to have such a file system.

If you accomplish your first goal, then you could concievably levarage that storage area as a source of your weekly offsite backup. Simply copy your storage area to your offsite media, and flush the history.


Hope this helps.. Good luck

 

[Jim_L9]

If you can initially sync up the 2 remote hard drives and keep them on a server somewhere else you can use rsync http://www.gaztronics.net/rsync.php to keep the data up to date on the remote HD on a server.

Problem is that there is no way to connect the PC with the server since the latter needs to be 10 miles away, and the only option is a very slow ADSL.

That is fine because rsync only copies changed blocks. If you make a small change to a huge file you don't resend the entire file. Much more efficient than copying the entire file for each change. You can also exclude copying your pagefile, etc.

 

[CaptRobertApril]

Yes, it was very helpful, thank you. The Rocket Raid basically incorporates what I was discussing as "floppy-net." The interesting part about it is that it has external boxes that could be duplicated and one set always left with the system.

Therefore, if I'm reading this correctly, the situation would be as follows:

Inside the Tower Case:
1 x 160GB Raptor
2 x 750GB RAID 1

In External Case 1
2 x 750GB

In External Case 2
2 x 750GB

At any given time one of the External Cases is connected to the PC and the other case is offsite. Now excuse my severe obtuseness (is that a word?) but in what RAID config (if any) would the External Case drives be set up as? Would I have to fully overwrite both HDs in each case weekly? Your description of stripes and mirrors was very clear but my inherent lack of education in RAIDing is making my freakin' head spin.

Let's keep in mind: Downtime is not a huge factor, data security is. Therefore the risk of thrashing even one file must be kept to an absolute minimum. Simplicity has to be key.

 

[lucas_donaldson]

Hey man the best backup tool I've seen is Symantec Livestate Recovery. We have many customers using this tool and it works great. A little pricey but its better than losing everything. You can schedule all you backups and push them to network locations without disturbing the user at all.

 

[pkquat]

I think a live backup program would be your best bet, and use a set of drives for offsite storage. A place where I worked had a similar system. Sorry I don't know what it was, I am not an IT person. This system would do incremental backups to HD's 2 or 3 times a day I think and it kept restore points for a few days. Each night there were incremental backups to tapes stored offsite that cycled through every week or maybe two weeks. Full backups were done once a month.

They used incremental because full backups were extra wear and tear on the equipment.

The server and the HD backups were set up as RAID 1 for up time reasons and for speed. The server had 3 drives in Raid 1 again for speed and uptime. Since most often data is being read from the server RAID 1 is more or less as fast as RAID 0 in this case. RAID 1 is only slower in writes.

 

[gse1]

In the neighbourhood where this PC lives a dude coming home from a bar last weekend was shot three times in the forehead for a few bucks and his Nike Airs. A $10K PC can be fenced for two grand and that will buy a good long fix for some horse addict. That's why the BU disks live ten miles away. If some bugeyed strungout freak breaks in, I want to hand him the PC, wish him the best of luck, show him the door and call my insurance agent without the added ballast of a few lead bullets in my brain.

Yikes, nice area, and I now totally see your thinking! Off-site backups are a requirement alright!

That is fine because rsync only copies changed blocks. If you make a small change to a huge file you don't resend the entire file. Much more efficient than copying the entire file for each change. You can also exclude copying your pagefile, etc.

It's only fine if you are making small changes in between rsyncs. CaptRobertApril has explained:

The number of apps I'll run in the average week can run up to 20, including 15 brand new ones. Everything from image editing to massive CAD files to ??? and lots and lots and lots of video, including some hidef.

This is too much to be rsyncing over an ADSL line as changes in 24 hours could regularly be gigabytes. External removable disks are the best answer (make sure you get USB2 ones too )

 

[CaptRobertApril]

In the neighbourhood where this PC lives a dude coming home from a bar last weekend was shot three times in the forehead for a few bucks and his Nike Airs. A $10K PC can be fenced for two grand and that will buy a good long fix for some horse addict. That's why the BU disks live ten miles away. If some bugeyed strungout freak breaks in, I want to hand him the PC, wish him the best of luck, show him the door and call my insurance agent without the added ballast of a few lead bullets in my brain.

Yikes, nice area, and I now totally see your thinking! Off-site backups are a requirement alright!

That is fine because rsync only copies changed blocks. If you make a small change to a huge file you don't resend the entire file. Much more efficient than copying the entire file for each change. You can also exclude copying your pagefile, etc.

It's only fine if you are making small changes in between rsyncs. CaptRobertApril has explained:

The number of apps I'll run in the average week can run up to 20, including 15 brand new ones. Everything from image editing to massive CAD files to ??? and lots and lots and lots of video, including some hidef.

This is too much to be rsyncing over an ADSL line as changes in 24 hours could regularly be gigabytes. External removable disks are the best answer (make sure you get USB2 ones too )

It's a lovely neighbourhood. The night is alight with the sound of gunshots and the screeches of the stabbed... It was recently suggested that the local buses add a direct route to the County Morgue. :lol:

USB 2 is the way to go on external disks. Definitely ditto there. The best (only) ADSL connection available here is a 256K up line. Best sustained up rate I've ever had has been about 30KBps. Given that on one weekly backup I've had to rewrite over 300GBs of new data, I don't think that online backup's gonna fly.

 

[ranger90]

...
Therefore, if I'm reading this correctly, the situation would be as follows:

Inside the Tower Case:
1 x 160GB Raptor
2 x 750GB RAID 1

In External Case 1
2 x 750GB

In External Case 2
2 x 750GB

At any given time one of the External Cases is connected to the PC and the other case is offsite. Now excuse my severe obtuseness (is that a word?) but in what RAID config (if any) would the External Case drives be set up as?

...

The hardware setup is spot on.

As for the RAID setup. You will want each RAID 1 (mirror) pair consisting of 1 internal disk, and 1 external disk. This way, you always have an internal copy, and an external copy. When you set all this up, be consistent. For example. Use the same slots in each external case. Don't put a disk in slot 1&2 in external case 1, and then use slot 3&4 in external case 2. When you set up your RAID1s. Be consistent there as well select the 1st internal drive listed, and the 1st external drive listed. Stuff like. You don't want to flip flop stuff around. It bite you in the buttox later.

For the weekly swap. You will want to "rebuild" your mirrors using the internal disk as the source of your material. That's under normal circumstances. If on the otherhand, the worst has happened, and you have replaced your machine and are restoring from your offsite, you will want to do a rebuild, using the external as your source. A "rebuild" or "copy" will start copying, block-for-block, every bit of data from source to target disk. Be prepared to wait a while.

Whenever you do your swapping, you'll have to "break the mirror", then disconnect the external drives. Then connect the "new" drives and rebuild the mirror. You will need the read your documentation carefully. Everybody uses different terminology.

Now for a big question. Is there an absolute need that these two 750GB drives be striped together in 1 massive 1.5TB (approx) disk? It adds a layer of complexity. You can do this, of course, And with an external setup like we've come up with so far, it might not be too bad. But if you accidently flip-flop drives and/or source/target destinations you could potentially destroy all your data on all your drives.

Considering you're just starting out. I'd stay away from adding that level of complexity until you're more familiar with your hardware and what it can and can't do.

I believe one of RocketRAID solution had room for 4 drives in each external case, and I believe 4 internal. You could always start out with 2 & 2, and once you're familiar with the hardware, and what you've done, you can do 4 & 4 and do bigger raid sets.


Remember document, document, document. Label, label, label. Even take pictures of your external cabling. There might be a chance of plugging cable A into slot B, and cable B into slot A. I don't know the hardware but from the beginning, think of these potential disaster points, document, and label them. And you should be able to get something working out...

Hope this helps.

 

[gse1]

As for the RAID setup. You will want each RAID 1 (mirror) pair consisting of 1 internal disk, and 1 external disk.

Ranger90, have you actually tested a configuration like this? My use of RAID is limited to a standard enterprise environment so I've never tried to do clever things like this, but I reckon any RAID controllers I've used so far (Adaptec and LSI Logic) would get upset pretty quickly if you start swapping around external disks in a RAID array. The problem is that some controllers store RAID information on the disks themselves.

However, I've only ever seen the behaviour of standard internal static RAID arrays so I may well be talking hot air. I suppose the right kit is out there that will let you swap around external drives, no idea what devices let you do this though.

 

[CaptRobertApril]

Thanks for the info and most of all for the warnings, Ranger. Yes, I can see how you could easily get distracted, plug cable A into socket B and end up with a foamy datashake.

Gse, given that the RocketRaid comes with it's own external boxes you would think that they had thought of what happens during swapouts and ensured that RAID info was written onto the HDs in such a way that they could easily be swapped, but then again, given the technological boners I've seen in my life, that is not necessarily a guarantee.