[Resolved]Greyhole, SnapRAID-One big array storage possible?

You can easily use greyhole+snapraid. I've been doing that for a year or 2. I use greyhole simply for pooling and snapraid for "duplication". If your using snapraid, you do NOT need to use greyholes duplication (ie, set the number of copies).

NeurOSick, I don't have my OMV interface in front of me at the moment... but something like:

(note, this is all of the top of my head)

* create SMB shares (when you do this you specify the hard drive where your "landing zone" (lz) will be... read about what it is here: https://wiki.amahi.org/index.php/Greyhole_landing_zone and here: https://github.com/gboudreau/Greyhole/wiki/AboutLZ).
* setup a greyhole pool and add your other drives to your pool, then add your pool to your SMB share (I think this is how it works, I'll have to look when I get home). (since your using snapraid as well... set the number of copies to 1)

You can setup as many SMB shares as you wish. All of those shares can use your drives... each one would work like a large pool.

Just remember, if you want to access your pool (your "one large drive") from OMV server, you need to mount samba locally on that server. It seems weird... but do it. You don't want to write directly to your greyhole DRIVES, you want to write to the mounted SMB share. Here is a script that does that very well: https://github.com/gboudreau/G…e/wiki/MountSharesLocally

No prob. I'm going to be doing a full writeup soon. The most confusing part of it is the "mount shares locally" thing. It seems weird to people at first.

I personally like greyhole a lot. It has always worked well and I've not had issues with it. I can't say the same for aufs (see http://zackreed.me/articles/84…view=comments#comment_216 for some aufs querks).

I've also not had any performance issues w/ greyhole, which make sense because its basically just reading/writing from a single drive without much overhead (just samba overhead mostly). I can max out my home gigabit (90MB/s reads, 75MB/s writes... which is pretty good on cheap consumer equipment... and my system is virtualized in xenserver) and can max out my drive speed if writing locally.

Zitat von "ryecoaaron"

I don't have any of those problems. Sounds like create=pmfs wasn't used and udba was set to reval. If you used the pmfs option and set udba to notify, I'd be curious to see if the problems still exist.

That being said, I usually recommend it for media collections. If you are creating/deleting directories that much, you should be using raid 5 instead.

I usually don't recommend raid 5 to people home since they don't usually have a UPS and there is usually not point in spinning all the drives and killing them quicker.

As for AUFS, I mostly started staying away from it b/c of custom kernal tweaks, exclusion from future kernels and such (not a for sure thing). Greyhole just works for me and was simple.

Zitat von "ryecoaaron"

How often do see things removed from the linux kernel? I really don't think aufs is going to be removed. And as long as you don't explicitly disable aufs in your kernel config, your custom kernels shouldn't have a problem either.

I like Greyhole (wouldn't have ported and added to the plugin if I didn't ) but I think AUFS is better for large, relatively static media collections.

AUFS (with NFS export enabled) is already a pain to install in lots of other OS's (including 12.04).

Also, large static media collections is what greyhole was made for. So it works very well.

Zitat von "ryecoaaron"

Not having a UPS may be typical for the average home user but I think they are much more common with people creating their own NAS. We definitely recommend them with OMV - Raid 5 or not.

In my opinion, starting/stopping the drive often is harder on them than idling especially if you are using a 24/7 NAS-rated drive. Also, read/writes contribute to wear and drives in a raid 5 array have less read/writes the more drives you add (obviously not with the parity drive though).

That is false. Raid 5 distributes the parity over all drives, not just one dedicated drive! Raid 4 has a dedicated parity drive, see https://upload.wikimedia.org/w…a/commons/6/64/RAID_5.svg and https://upload.wikimedia.org/w…a/commons/a/ad/RAID_4.svg.

Zitat von "ryecoaaron"

How often do see things removed from the linux kernel? I really don't think aufs is going to be removed. And as long as you don't explicitly disable aufs in your kernel config, your custom kernels shouldn't have a problem either.

I like Greyhole (wouldn't have ported and added to the plugin if I didn't ) but I think AUFS is better for large, relatively static media collections.

I doubt that aufs will be removed from the kernel, too. Linux Containers heavily rely on it.

Greetings
David