Hi,

 

I've been looking into building my own UnRAID server to take on all my network storage requirements.

 

One is that I can Time Machine my Macs to it.

 

As the TM backup spans more than the capacity of a single disk I need this user shares thing to work.

 

Does Time Machine work reliably or not?

 

If it doesn't, then I think I will have to ditch my plans as TM working reliably is my number 1 requirement.

I don't think you can span disks due to the way Time Machine works.

Yes, I've read it's best to designate a single drive for just Time Machine, as it's mounted as a disk, and will not cooperate with your backups scattered amongst multiple drives.

 

There are some posts in this forum discussing this as well.

In that case I will have to ditch it then as im currently spanning 2x2TB in a FW800 enclosure.

 

How do others use time machine if it's limited to the biggest HD available?

 

I have 4TB in my Mac Pro... find me a 4TB HD!

 

Any ideas if/when this feature will be available?

 

Edit: Actually, given that 99% of that 4TB isn't actually changing, I could possibly use a different backup method for the photo library and only use Time Machine on things that change regularly (Documents, Boot drive, etc). Brings it down to about 500Gb then.

 

If I assign a 2TB disk to time machine, will that solitary disk also be 'backed up' onto the parity disk?

If the disk is in your array, then yes, it will be included on your parity disk.

 

I just choose what I want to Backup, like Documents, possibly Preferences--anything that will actually change, like you have mentioned.

 

Maybe once a week/month, you could do a Carbon Copy Cloner of your OS partition so in case something happens to your OS drive, you can restore a fairly recent back up of your drive from the CCC image.

 

I don't see much need to use Time Machine to back up everything, but only essential things that I wouldn't be able to recover/re-download.

 

I'll probably dedicate 1TB to my Mac's Time Machine, as I only have a 320GB internal drive.  I wonder if I can actually use less of one drive, so maybe just limit it to 500GB of a 1TB.

I don't understand why TM would know that a user share is comprised of more than one disk. TM writes to a sparsebundle and the bundle grows in 8MB chunks. Each chunk looks to unRAID as a single file. The bundle contains an HFS+ file system so how does TM know or care that the user share is made up of more than one disk? UnRAID makes the multiple disks appear as a single volume.

 

I am running a test currently with a user share comprised of 2 disks set to most free allocation. I have two machines running TM backups to the share. It has been running well for almost a week. I plan to report in a months time if it still runs without issue. It will take at least six months to determine if it is stable enough.

I don't understand why TM would know that a user share is comprised of more than one disk. TM writes to a sparsebundle and the bundle grows in 8MB chunks. Each chunk looks to unRAID as a single file. The bundle contains an HFS+ file system so how does TM know or care that the user share is made up of more than one disk? UnRAID makes the multiple disks appear as a single volume.

 

I am running a test currently with a user share comprised of 2 disks set to most free allocation. I have two machines running TM backups to the share. It has been running well for almost a week. I plan to report in a months time if it still runs without issue. It will take at least six months to determine if it is stable enough.

 

In that configuration however are the files on the disk directly accessible?

 

Normally, as I understand it if 1 out of 2 drives in a user share dies, you can still access whatever files there are on disk 2.

 

Now if TM makes an image, most of the files are just links to other files, meaning you could just have a disk full of symbolic links and not actually the files, which defeats the point of this kind of secondary 'safety' net.

 

The more I think about it, the less attractive spanning TM over two disks is!

I could test with some 160gb drives and see what happens,  I'm pretty sure someone has already reported issues with spanning disks in another thread however.

I don't understand why TM would know that a user share is comprised of more than one disk. TM writes to a sparsebundle and the bundle grows in 8MB chunks. Each chunk looks to unRAID as a single file. The bundle contains an HFS+ file system so how does TM know or care that the user share is made up of more than one disk? UnRAID makes the multiple disks appear as a single volume.

 

I am running a test currently with a user share comprised of 2 disks set to most free allocation. I have two machines running TM backups to the share. It has been running well for almost a week. I plan to report in a months time if it still runs without issue. It will take at least six months to determine if it is stable enough.

 

In that configuration however are the files on the disk directly accessible?

The mac files are not directly accessible on unRAID. The mac files are stored in the sparcebundle and are accessible once the spacebundle is mounted on OS X and via TM. The sparcebundle chunks are in the sparcebundle which is directory as far as unRAID is concerned. Each chunk is an individual file within the sparcebundle/directory. I have been splitting the sparcebundle diretory with success so far.

Normally, as I understand it if 1 out of 2 drives in a user share dies, you can still access whatever files there are on disk 2.

Normally you can access all of the files, including those on the failed disk, due to the parity protection providing emulation of the failed disk. The failed disk can be rebuilt with continuous read/write access to the emulated disk.

Now if TM makes an image, most of the files are just links to other files, meaning you could just have a disk full of symbolic links and not actually the files, which defeats the point of this kind of secondary 'safety' net.

TM uses a sparcebundle not an image. The sparcebundle is a directory filled with 8MB files (the chunks). The symbolic links occur in the HFS+ files system that resides within the sparcebundles. There are 2 layers here. TM sees the sparcebundle as an HFS+ disk. UnRAID sees the sparcebundle as a single directory containing 8MB files.

The more I think about it, the less attractive spanning TM over two disks is!

 

Totally disagree. As long as all of the component 8MB files/chunks are presented by unRAID the sparcebundle should operate correctly.

 

The more I think about it, the less attractive spanning TM over two disks is!

 

Totally disagree. As long as all of the component 8MB files/chunks are presented by unRAID the sparcebundle should operate correctly.

 

 

Well from a parity disk failure PoV if another disk dies in the TM User share you are ****ed.

 

Where-as normally you would be at-least able to get the data off the disks which haven't failed.

 

In my opinion from a safety net of not loosing everything, I will not be spanning TM disks. Besides I don't actually need to, I'm going to be using another way to backup the photos/videos from my DSLRs.

 

 

The more I think about it, the less attractive spanning TM over two disks is!

 

Totally disagree. As long as all of the component 8MB files/chunks are presented by unRAID the sparcebundle should operate correctly.

 

 

Well from a parity disk failure PoV if another disk dies in the TM User share you are ****ed.

 

Where-as normally you would be at-least able to get the data off the disks which haven't failed.

 

In my opinion from a safety net of not loosing everything, I will not be spanning TM disks. Besides I don't actually need to, I'm going to be using another way to backup the photos/videos from my DSLRs.

 

 

 

If your spanning more than 2 disks and 2 of the spanned disks fail concurrently then your right. If you span only 2 disks then there is no additional risk: if one fails it can be recovered and if both fail then you've lost only the contents of those two disks.