Here is my situation:

I have some digital data for which I need high redundancy but not necessarily high availability so I am trying to cover both the scenario of disk failure and that of NAS failure.

I was thinking of purchasing a two-bay NAS which uses the EXT3 filesystem, and setting up a (software) RAID 1 volume on it, and using an eSATA connected drive as a backup (also using EXT3).

I know that a RAID 1 volume will continue to operate with one disk failure until replacement of the failed disk and rebuild of the array, thus covering the scenario of hard disk failure, but what about a hardware failure of the NAS itself? Would it be possible to remove either one of the disks from the NAS and connect it to a Linux PC and access the data since the disks are simply a mirror of each other with no parity data?

For a NAS using software RAID 1 and EXT3, what would be the likelihood of data corruption assuming NAS hardware failure?

Are there any better alternatives such as a multi-bay NAS using RAID 5 or 6? I assume that once the RAID level includes parity data, it would not be possible to simply pull out one of the disks and read it in a PC.

If it really is just plain EXT3 and RAID1 then yes, you can just take either drive and insert it into any Linux box and access it.

Don't forget that RAID1 does not equal two discs. You can have as many discs as you want in you RAID1 array, adding more hardware redundancy.

And yes, once RAID5/6 are in play you can't take single discs out and take your data elsewhere. I a RAID5 array with N discs, you need N-1 discs from the array to recover your data.

However if you are simply after redundancy they you might be just as well off with a regular rsync job that synced between two Linux based single drive NASes.

It's sometimes tricky though to identify what NAS or RAID devices will properly support RAID 1 across more then 2 drives, vs a RAID 1+0 or 0+1 setup.

Peter, do you plan on using a Linux machine to read the disks at some point, or an ext3 driver for another OS? In the case of the ReadyNAS devices, the data should be readable when attached to a Linux machine with LVM support. Reason I bring it up is that some of the ext3 drivers for other OSes can't read partitions inside an LVM container.

I wonder if there are any cheap single drive NASes out there that have builtin support for syncing with a second NAS ?

I could use either an ext3 driver for OSX or Linux in a VM (which would be my preferred solution) but these are both a last resort for retrieving current data assuming a hardware failure has occurred prior to the next scheduled backup to the external eSATA drive. I am not certain whether the devices I am considering use partitions inside a LVM but if I use Linux in a VM then that should not be an issue?



The devices I am looking at (Synology) are not the cheapest but they do specify support for backups over the network to another Synology product or other rsync compliant NAS. I believe this is a 'point in time' backup though, rather than a continuous realtime mirror like RAID 1. I will still have a once-daily backup to an external eSATA drive, plus a manually initiated rsync backup to a 2.5" portable USB HD for offsite backup.

Basically, I am trying to cover myself against disk failure in the NAS and / or NAS hardware failure as a worst case scenario.

Be aware that direct disk access in a number of VM programs (VMware, whatever MS VM is called and a couple of others that I can't remember) is (at best) difficult to use particularly when you get to drives over 1TB.

What I was trying to do.
http://lime-technology.com/forum/index.php?topic=5030.0
http://communities.vmware.com/message/1448042#1448042

I did get it working with a special SCSI driver in unRAID eventually but had to resort to windows command line (DISKPART) to make disks go offline properly.

Thanks for the heads-up on that Shonky.

By direct disk access do you mean like a block copy? I was hoping that if I run into the situation where I need to pull a drive out of the NAS, I could use something like Ubuntu under Parallels on the Mac (with the disk connected in a dock via USB) and simply mount the disk and copy files off it, or if necessary I have an older WinXP desktop I can connect the drive to internally via SATA and dual boot into Ubuntu.

He means mounting the real physical disk into Linux running in a vm, as opposed to mounting a virtual disc image like you typically do with a vm. You'd probably find that task a lot easier under a live CD booted on the PC than under VMWare on the Mac.

Surely this is in fact the whole problem -- i.e. whether the partitions are "bare" on the drives, or whether there's some sort of metadata stored by the NAS that offsets the partition table and makes the partitions hard to mount on other systems. I don't know whether NAS boxes in general have metadata "partitions" before the real data, let alone how to tell whether any particular one has.

Peter

If you create a filesystem on a partition on the disc, you will need to give the VM direct access to the disk so it can see and mount that filesystem. The recommended method for disks in a VM these days is to create a disk image file on a partition mounted in the host system. Direct disk access seems to have largely fallen out of favour. If ext3 there are some Windows drivers I've seen but when I've tried to use them they are not the most reliable so I've resorted to a live CD (as mentioned below).



Yep. And what you're suggesting Andy would work just fine and would be easier for a one off occasion where you're basically trying to recover from a failure. Surely there are live CDs that would boot on a Mac these days?

I believe there is an operating system partition and a swap partition before the data partitions on each disk in a Synology NAS (created during the firmware installation) but they should be part of the partition table AFAIK.

I got the information that they run ext3 from here and apparently the latest version of Synology's NAS firmware, DSM3.0, gives the option of using ext4 filesystem but no mention of anything non-standard about the partitions.

I would anticipate that the Synology Hybrid Raid (from the second link above) would not be able to be read outside of the NAS itself but that is only speculation.

Consider humans when designing any backup scheme. People will delete entire filesystems by accident, and any raid solution will immediately replicate the deletion. They'll have deleted a file a month ago and want it back. They'll have overwritten a file with bad data and want the old file back from last week. This human may be you.

Hardware failure is real, but it's probably the least of your worries if you're discussing backups and not availability.

Excellent point, Matthew.

RAID isn't a backup. But I'm sure everyone here knows that already.

All good points Matthew; I imagine that a Time Machine type of setup would be ideal, however I don't know whether that works over the network.

After some googling, it appears as though these guys reckon it is possible to take the disk(s) out of the NAS and read data directly from them.