Mhddfs help

RJR · 3. Januar 2015

I have just moved my NAS over to OMV and have everything setup with the exception of Mhddfs. In my old system I had a folder called /mnt/Movies which was shared and the Mhddfs pool was pointed to individual folders on each drive such as /mnt/Disk01/Movies with only one drive as rw and all others as ro.

I cannot quite figure out how to do this within the plugin as it only gives choices of devices. Is there a configuration file in use for Mhddfs so that I can simply add my own commands such as rw/ro options and point to specific locations?

RJR · 3. Januar 2015

So I figured out some portion of this such as the folder options, it seams that even when you try and point the config.xml file to a deviceid/folder it will not allow it as it wants all the folders to be named the same. This is not that big of a deal as renaming about 13 folders out of 22 is much better than moving around a whole lot of files.

The only thing I need to figure out now is how to get the system to only write to one specific drive if possible. Not sure how I got this done last time but with my last system when I downloaded fan art or subtitle files or created an nfo file they were always kept on an ssd drive so that when you browse for movies all of those files were available without spinning up the entire drive array. This always allowed KODI clients to load up and browse all of my movie collection without spinning up all the drives.

Any thoughts would be helpfull

mkgriesinger · 3. Januar 2015

I am not sure MHDDFs has this granularity. From what i understand, it writes to a disk until its full, then moves to the next disk. AUFS will "Stipe" the files across all the disks in the pool, but writing the data to the disk with the most free space.

I learned some tricks for mhddfs from the following website

http://svn.uvw.ru/mhddfs/trunk/README

From /etc/fstab
mhddfs#/media/SDA,/media/SDB /media/Pool fuse defaults,nonempty,allow_other,mlimit=15G 0 0

I specified my system to move to the next disk once there is only 15GB of free-space left by adding mlimit=15G.

ryecoaaron · 3. Januar 2015

Using the whole drive is what mhddfs does in its space free calculations whether you use folders or the whole drive. You should create a mhddfs pool and share it as RW using a shared folder with those permissions. You should not share the other drives because there really is no reason too. If you want to share them as RO, then create a shared folder for each drive with read only permissions.

The mlimit option will be overwritten if you change anything, mhddfs defaults to 4gb so we didn't make it an option to change in the unionfilesystems plugin.

TheOneAndOnlyJH · 28. Februar 2015

I liked the way the older plugin used shares for branches rather than the whole disc. I've looked at using the new unionfilesystems plugin but I would have to reorganize my entire shared folder structure and permissions to get it the way I had things before.

@ryecoaaron - Is the older standalone mhddfs plugin still available for download? (I believe I originally found it on GitHub, but I can't locate it anymore.) If you have a copy would you mind sharing it?

ryecoaaron · 2. März 2015

link

BigBoo · 26. März 2015

Zitat von ryecoaaron

Using the whole drive is what mhddfs does in its space free calculations whether you use folders or the whole drive. You should create a mhddfs pool and share it as RW using a shared folder with those permissions. You should not share the other drives because there really is no reason too. If you want to share them as RO, then create a shared folder for each drive with read only permissions.

The mlimit option will be overwritten if you change anything, mhddfs defaults to 4gb so we didn't make it an option to change in the unionfilesystems plugin.

I'm a recent unRaid convert and in the process of moving my data to OMV. I've been able to work around any issues thus far and thank all those involved. I've hit a bit of an problem with my setup using mhddfs, xfs and snapraid - and at the moment I am looking at the mlimit option so I hope my post is relevant to this thread. I'm running 1 - 4TB parity, 2 - 4TB data disks in a mhddfs pool. If there is no possibility to enable mlimit option, as it works with my setup, the write process fills disk1 to capacity as expected before moving to the next disk. (4GB reserve but only when a new write calculation is performed. If writing has established prior it can write into the 4GB leaving very limited space.. a benefit with xfs). However to snapraid sync a 4TB disk the parity must be able to store a little more than the data disk (equivalent of 50GB reserve on data disks for that 4TB parity) and so the sync causes an error: not enough space on parity, free up space, delete xyz files, etc. Even if I move data off that disk1 to free the approx 50GB needed to run a sync, the next disk write to the pool will fill that disk back up and next sync will fail.

I'm running the xfs system due to a disk sleep issue that was showing with ext4. I know that in ext4 I can tune2fs and shrink the data disks to reserve space (root space) but there is no such option with xfs file systems as it goes against the inherent intent of fully utilizing all space.

I assume that any change to the mhddfs pool when I add my next set of hard drives would put the 4GB lmit back in if I set things manually. Is there another option here? Would it make sense to request a feature for this? If using xfs, mhddfs and snapraid enable an option to change the defaults?

Thanks in advance.

ryecoaaron · 26. März 2015

I will look into adding mlimit.

BigBoo · 26. März 2015

Awesome! Thanks ryecoaaron.

ryecoaaron · 26. März 2015

version 1.4 has mlimit (threshold) for mhddfs and is in the repo now

BigBoo · 26. März 2015

wow.. Super FAST! Thanks again.

alxscott · 26. März 2015

First post!!

Looking to move from unRAID to OMV; for the same reasons already stated plenty times in the forums!

I've got 10tb on 4 x ReiserFS disks from unRAID and want to mount and use 'as is' until I replace the drives; to that end I'm running an OMV VM to see my options. I'm happy that the main file system page doesn't really like ReiserFS, so I have mounted them manually. Because I'm retaining the original folder structure; I'm manually adding the drives to fstab; with the allow_other and nonempty options.

Down to my point; currently in the UnionFS plugin page it only shows the EXT4 drives I've added ( to see if I can format my new drives EXT4 and use with the old ReiserFS rives in a pool). Would it be possible to:

a) Add drives which are manually mounted (i.e. not via the WebGUI as this does not support ReiserFS) .

b) Add the "nonempty" option for drives with existing folder structure?

Apologies if this has been covered before; been Googling unRAID -> OMV migration but not seen it

Thanks in advance

Alex

ryecoaaron · 26. März 2015

The unionfs plugin only shows filesystems mounted using the OMV web interface. You will find this concept throughout the plugins. Even though you can't create a reiserfs filesystem in the web interface, you should be able to mount it in the filesystems tab. This will add the entry to fstab and allow you to use it in plugins.

What does the nonempty option do?

alxscott · 27. März 2015

Thanks for you quick reply!!

If i try and mount the volume in the WebGUI I get an error ( Screenshots attached). Tho this might be because its a mock up in a VM; I haven't tried with my actual data disks yet!

The nonempty, I belive, allows you to mound hard drives with existing data. It won't let you mount without it. Again, screen shots attached

Alex

ryecoaaron · 27. März 2015

I would file a bug report here. There is a backend for reiserfs. So, you shouldn't get that error.

I remember what nonempty is. It for when you mount a drive/partition over a directory with data in it. You shouldn't have any data in the /media folder on the OS drive. So, you don't need nonempty.

Jetzt mitmachen!