Found this, looks really good, includes inline deduplication and all other ZFS features, but easy to use as an appliance: http://sites.google.com/site/eonstorage/

It's all good, but when I watch a movie eon would need all my 20 disks spinning.

UnRAID would need only one disk spinning.  That's the deal-breaker for me.

 

you also cant expand a z pool very easy (keeping data intact). with unraid you add a disk, clear it, pool size goes up. with riad z you add a disk, delete your old array, make a new one with the extra disk, copy crap back to it all over again

 

easy network teaming is something that unraid needs though.

Terrastrife, you expand the set online by adding raidz or raidz2 or raidz3 pools to it. each raidz consists of say 3 or 4 disks so that this is the upgrade step. Adding raidz' increases performance as data is sriped accross them. No rebuilds necessary while upgraing and the array is protected all the time.

 

However, as Purko said spinning only a single disk when others aren't needed is a deal breaker!

so it supports online expansion then? so say i have 5 disks in a raidz i can add a 6th to the same array?

 

because this was the only thing stopping me going nexentastor vs unraid. i dont want multiple pools, i want one pool, one array. dont want to waste capcity on parity.

online but not to the raidz, only to the volume. A volume is compised of at least one raidz or a lot of them (and data is being striped between them).

 

Actually everything is online without rebuild, you can add single disks to mirror/stripe/concat'ed volumes, but not to raidz/2/3 sets because they are like raid5/6/7 and ZFS do not rebuild bud aims to be asap available with increased and not decreased performance + risk during a complete subsequential raid5 rebuild lasts.

 

So, you can have one volume with a lot of raidz sets and each raidz set would be typically 3..8 disks with one or more disks for parity. The full amount of disks for a set would be needed when an (online) upgrade is to be performed. You can mix different set widths and raidz-levels into a volume.

 

From an availability point of view a 6+2 raidz2 set would be better then two 3+1 raidz sets, but then the upgrade step would be 8 disks at a time. And yes, you loose much more on parity, but you gain much more performance :-)

yup, and thats why i ddint go zfs.

 

a real raid5 offers online expansion and full protection during the expansion. i know it is possible to expand a raidz but reading up it is a massive task, even for someone whos pretty good with linux and thers a high risk of data loss.

 

performance depends on how you use it. if you have n aggregate 4 gig link for example, then yes a single process will go super quick, but for msot of us teaming means load balancing which is relly 4 single gig links so there sno reall performance gains...if unraid could team, that is.

ZFS wasn't designed for raid5; it was designed for striping and mirring and to be used with 10GigE interfaces :-) Its a Zettabyte FS on a 64bit OS. Raidz came in later and then raidz2 and raidz3.

 

Edit: a raid5 can loose data, read on raid5 write hole. ZFS fully eliminates this and adds other safety features, it is the most safer FS to date.

And I am not talking about raid5 controller failures or controllers without mirrored cache (hardly anything that costs less then 5 digits) and the insanely increased risk of loosing data while expanding raid5 volumes. Especially with todays large disks you would be running absolutely unprotected and fully stressed for at least 10-20 hrs. Handling expansion differently was the 2nd feature unRAID caught me on in addition to single disk spindown.

raid5 expdands will full protection. actually even if youre changing raid levels yuo maintain the lowest level of protection. so expanding raid5 to raid5 you get raid5 protection, or raid5 to raid6 you get raid5 protection during the expansion.

raid5 expdands will full protection. actually even if youre changing raid levels yuo maintain the lowest level of protection. so expanding raid5 to raid5 you get raid5 protection, or raid5 to raid6 you get raid5 protection during the expansion.

 

During raid5 expansion you loose *absolutely* ALL protection, this is why there is raid6.

raid5 expdands will full protection. actually even if youre changing raid levels yuo maintain the lowest level of protection. so expanding raid5 to raid5 you get raid5 protection, or raid5 to raid6 you get raid5 protection during the expansion.

 

During raid5 expansion you loose *absolutely* ALL protection, this is why there is raid6.

 

my raid controller does protected expansion just fine. power outs arent a problem even without battery backup or ups, and stopping / resuming isnt an issue either.

 

most ppl still belive raid5 is like it was 15 years ago. controllers have come a long way. no more failure  with errors on rebuilds etc these days.

most ppl still belive raid5 is like it was 15 years ago. controllers have come a long way. no more failure  with errors on rebuilds etc these days.

 

They have come a long way, yes, but there are a lot of *potential* safety issues, raid write hole or component failures.. thus high-end storage arrays use all redundant controllers, two in parallel, with mirrored and battery backed up cache, etc to *make* sure noting can happen while online rebuilding or expanding or migrating raid arrays. Just read up on the net. Disconnect the power, some chip or memory fails (on your $1000 home raid controller) while expanding and all your valueable data is dust.

sure i may lose s small amount of data, and result in a missing frame or two in a video file, but this is fine for most home users (with no battery backup).

 

raid is not a backup solution remember.

You loose ALL the array.

You loose ALL the array.

 

no, ive had an error durig an array rebuild, you only lose the that particular sector (i have 10.5 and 7 TB raid5's on my desktop).

 

like i said, raid5 has come  a long way the last 15 years, a error on rebuild does nto drop the whole array anymore, only the data on that part of teh stipe is gone. best to check yoru log if you let it continue on errors though. its an option in most raid configs.

You loose ALL the array.

 

no, ive had an error durig an array rebuild, you only lose the that particular sector (i have 10.5 and 7 TB raid5's on my desktop).

 

like i said, raid5 has come  a long way the last 15 years, a error on rebuild does nto drop the whole array anymore, only the data on that part of teh stipe is gone. best to check yoru log if you let it continue on errors though. its an option in most raid configs.

 

This isn't my experience / is only half the truth.

 

If there's a transient read error you *might* have a rebuild but less that block(s) of data. And you *might* be lucky that the failed block isn't part of the filesystem meta data.

 

What is the bigger issue is a terminal read error on a disk during a rebuild. All the hardware array controllers I've experienced don't really treat things on a block by block basis. If you hit this kindof issue then they flag the disk as bad and pop it out the array.

 

As already said if this happens during a rebuild, you have no protection and your array has failed.

 

There may well be controllers which don't do this, but I'd make the argument that this is better than just blindly stumbling on and trying to reconstruct data anyway as there's a big chance you'd end up with silent corruption.

 

All the enterprise arrays I run, bar none, mark a disk as bad with relatively little reason. None try to continue to read, rebuild or perform any disk i/o on a potentially dubious disk. It's immediately ejected from the array.

 

I wouldn't want it any other way. RAID-6 and hot spares make this (almost!) a non-issue.

 

As you say, raid is not a backup and in many situations I'd rather have the whole array fail than have a big chunk of it corrupted (and overwriting good data during a backup cycle!) without me knowing.

You loose ALL the array.

 

no, ive had an error durig an array rebuild, you only lose the that particular sector (i have 10.5 and 7 TB raid5's on my desktop).

 

like i said, raid5 has come  a long way the last 15 years, a error on rebuild does nto drop the whole array anymore, only the data on that part of teh stipe is gone. best to check yoru log if you let it continue on errors though. its an option in most raid configs.

 

That is not just too simple. Here I've got a bad backplane, and the controlled dropped a disk. When the rebuilding started, a second disk was dropped too, and the array become out of sync. Therefore I would not be able to mount the array again. That's how I came to unRAID world. Its a very sad history with the lost of 5TB of valuable data. RAID5/6 are not suitable to home use, unless maybe for a few users. We do not have redundant PSU, ECC memory, a large SAN for full backup. We use consumer grade products, not server grade ones. In my case, the faulty backplane wans't a Supermicro one, but a Norco.

 

This is why, IMHO, striped data arrays are evil to home users.

when i drop two disks, my array goes offline. if it doesnt pick up teh disk again if its not a disk fault, i can reload the disk formatting/raid data. i have a highpoint card, hpt offer you a dos program to backup all your raid/striping data in case of failure in that parts. its not on the website, like a few other tools that are handy (like staggered spin tool etc) but you gotta email them.

 

but yes data currupt is bad, but having that is better then not having it at all.

 

also, i am moving to unraid, setting up a couple of sheap cases and old hardware so i can get rid of my raid5, which wasnt idea to begin with for what i want to do.

I've done testing in a VM environment on raid5 and raid6 expansion - to the point where I've powered off the VM in the middle of a raid5 expansion (to simulate loss of power).. The machine came back up and continued the expansion where it had left off.

 

No data was lost. I've done the same with RAID6.. same results.

 

I've done testing in a VM environment on raid5 and raid6 expansion - to the point where I've powered off the VM in the middle of a raid5 expansion (to simulate loss of power).. The machine came back up and continued the expansion where it had left off.

 

No data was lost. I've done the same with RAID6.. same results.

 

 

That's a different scenario than we are discussing. That's fairly standard / usual.

 

Start your expansion and mid way through power off your machine and remove a virtual disk / remove the virtual disk mid expansion. To simulate a disk failure during the expansion...

I must have read too quickly..

 

I'll add that to my list of things to try...