Big thanks for the fast reply! I overlooked the fact, that it was an Ubuntu-package.
I already had the testing repo activated, but I thought mergerfs would show up in the available plugins-list then, which it didn't, so I thought it wasn't available from OMV-Extras.
However, thanks to your help it's now working
Unionfilesystem Plugin
-
- OMV 1.0
- sno0k
-
-
Can you add the fstab options "nonempty" and "direct_io" to the mergerfs plugin ?
So its selectable in the unionfs plugin and nobody has to edit fstab manual.I ask because i added this options to my fstab and after todays update it got overwritten.
Thanks
Hey, what does "nonempty" do? I couldn't find it mentioned at all on https://github.com/trapexit/mergerfs .
-
nonempty is a standard fstab option (not just mergerfs) that allows you to mount a filesystem to a non-empty directory and be able to see both sets of files/directories.
-
-
Thanks! After your hint I searched for fstab nonempty but I couldn't find any explanation. Makes me wonder how that option works, because just by your explanation one could think it would render things like mergerfs obsolete, although of course I'm sure that isn't the case.
-
It definitely won't render any union filesystem obsolete. All it lets you do is put files in a directory on root (/media/test for instance) and the mount another filesystem at the same location and see both sets of files. You can't mount more than one filesystem at that spot. It really only "helps" if you write to the directory while the filesystem is not mounted (error?) then you mount the filesystem and hope to see the files you wrote to the root filesystem.
-
I got it Thanks!
-
-
Hello!
First of all a huge THUMBS UP for the OMV project! I'm in the process of rebuilding my little home NAS and want to switch from a plain Ubuntu to something related with a nice GUI.
Can you add the fstab options "nonempty" and "direct_io" to the mergerfs plugin ?
So its selectable in the unionfs plugin and nobody has to edit fstab manual.
I'd like to second this request!I'm planning to migrate from mdadm RAID-5 to SnapRAID and a Union FS. I first tried AUFS, but I don't like the mess it makes with its whiteout and opaque files (I prefer the data disks to stay clean of such stuff, in case I want to continue with another or without a union FS).
So my next choice was MergerFS. My NAS is a low-power machine (Intel Atom D510), and since a FUSE FS naturally has a much higher demand on the CPU, I found that using the "direct_io" option yields much better results for my system (write speed 120 MB/s as opposed to 40 MB/s, while not much impact on read speed).
Thanks in advance, and again thanks a lot to the team for your great work!
-
You can already do this by adding OMV_FSTAB_MNTOPS_MERGERFS="defaults,allow_other,direct_io" to /etc/default/openmediavault and then running omv-mkconf fstab.
-
You can already do this by adding <code class="inlineCode">OMV_FSTAB_MNTOPS_MERGERFS="defaults,allow_other,direct_io"</code> to <code class="inlineCode">/etc/default/openmediavault</code> and then running <code class="inlineCode">omv-mkconf fstab</code>.
Many thanks, that worked great! (As expected it also applies to newly created MergerFS filesystems.)
Is there maybe some kind of list of these entries one can add to the <code class="inlineCode">/etc/default/openmediavault</code> for "advanced config"?
-
-
Yup, you can find that under wiki.openmediavault.org (environment variables).
Greetings
David -
Anyone else besides me seeing crashes with mergerfs? I'm pretty sure it boils down to high I/O but need to gather some logs
https://github.com/trapexit/mergerfs/issues/188 (trapexit thinks it's related to libfuse or kernel as they're somewhat dated in Wheezy)
-
I'm using it on jessie and just transferred about 50gb last night using rsync with no problems. About 68 MB/s continuous.
-
-
-
I wanted to run it a bit longer before trying the directio flag. I just sync'd 140 gb worth of music files with no issues.
-
I run for 3 weeks and no problem detected, only speed improvement
-
-
Added the direct_io flag and writes went to 80 MB/s. Nice boost.
-
Yes, but reads really suffer. I'm pulling across a 10 Gig file at about 16 MB/s when direct_io is on, 85 MB/s when direct_io is not on.
It's my understanding that this is more or less expected because the point is caching, but it sure would be nice to have some way of automatically using direct_io on writes but not reads.
-
-
-
Cool, I missed that, thanks.
direct_io not enabled + write-cache disabled: read rate 85 MB/s
direct_io not enabled + write-cache enabled: read rate 70 MB/s (?)direct_io enabled + write-cache disabled: read rate 16 MB/s
direct_io enabled + write-cached enabled: read rate 21 MB/sI haven't looked at the write performance because I trust what others have seen; I'm just noticing a big hit on reads when direct_io enabled. I have no idea what's going on or even if it is unusual, nor am I complaining. These are just naive observations on my part.
Also I think mergerfs & OMV are tops!!
-
I just put mergerfs 2.9.1 in the repos. Maybe it will help performance. I updated but haven't tested performance yet.
Participate now!
Don’t have an account yet? Register yourself now and be a part of our community!