How to install and setup MergerFS

[@trapexit]

mergerfs author here. Thanks for covering software. It looks like you installed the Fedora 31 RPM. If it works it works but I should have all the recent Fedora and EL builds too.

[@distrodomain]

Oh that's great!, out of simplicity I installed FC31 package instead of compling from source, thank you for the great software, looking forward to EL9 build

[@trapexit]

@@distrodomain I have been reworking my build process and added RockyLinux 9 support so it will be in the next release (which should be coming out soon'ish.) I could add them for 2.35.1 if needed.

[@distrodomain]

Can't wait to read the changelog of the new release, I use MergerFS on all my servers at home.

[@trapexit]

@@distrodomain Glad to hear. Won't be anything too crazy.

[@noobstrom3279]

Can i use this method if i have file already on a drive?

[@vijinho]

Thank you for your hard work, your two videos have proved invaluable to my learning and are much appreciated. I am experimenting in switching from a very reliable ZFS, somewhat slow (83% used, 30% fragmented) now 7 x 3TB RAID6 home server with LSI 92xx SAS HBA to something like you're describing.

[@tandyman6672]

Excellent guide to installing mergerfs! I really appreciated your clear description of the actions needed and their purpose. ⭐⭐⭐⭐⭐

[@distrodomain]

I’m glad it was helpful, thanks for watching !

[@jlperformance1539]

Super helpful! Going to set up snapraid now, thanks to your videos!

[@distrodomain]

Glad to hear, make sure you don't miss the video about how to recover a drive with snapraid kzbin.info/www/bejne/nJ7YmoWChciogKc

[@jlperformance1539]

@distrodomain you ever consider doing a video on how to setup a network file share? I got it figured out but I'll be honest, I don't understand 90% of the stuff that I did to configure it. That's what I like about your videos, you explain how to configure it, and *why* you're configuring it that way. Helps to really understand the concepts at hand.

[@jlperformance1539]

@@distrodomain So after restarting my server, I had to go through and re mount all my drives. How do you automate this process? Did I miss something in the video?

[@distrodomain]

I'll make a video about shring this via samba or smb for windows and nfs for linux clients, It's a great idea, thank you for the comments.

[@distrodomain]

To automatically mount them, check that your fstab doesn't have typos and that the id of the disk matches your disks, by executing mount -a it will mount all mount points configured in the fstab and should confirm that the configuration is good, let me know how it goes! :)

[@-gorman-]

Thanks for this. After spending two hours banging my head against the wall, your very clear step by step helped me enormously. Liked and subscribed to the channel! 😃

[@distrodomain]

I'm glad you found it useful, thanks for watching! :)

[@clownshoe1413]

What a great tutorial. Explained perfectly. Liked and subscribed.

[@distrodomain]

Glad it was helpful!, thank you for watching :)

[@ShaferHart]

A lot of good tips in here. Thanks for sharing.

[@distrodomain]

Glad it was helpful!

[@YofuMedia]

amazing content!! thanks for sharing!!

[@distrodomain]

Thanks for watching!

[@YofuMedia]

@@distrodomain big fan !!!

[@MusicFluff]

Great video in combination with the Snapraid tutorial. A tutorial on setting up tiered caching with ssds would be great too!.

[@distrodomain]

I'll look into it, thank you for watching!

[@andresvidoza]

Love the video. I am also currently watching your Snapraid tutorial as well to get an overall understanding of how to protect and sort my data. I do have a question tho, is it possible to then use Samba to share these drives on the Local Network? I use drives to store footage. Thank you!

[@distrodomain]

Yes this is exactly what I do, I install samba and create a share, sharing the root mount point of mergerfs, you can control permissions and access with the smb.conf file. If you have any issues let me know I’ll be able to help you. Thank you for watching.

[@mrnobody7025]

This is a great video. Thank you!

[@distrodomain]

Glad you enjoyed it!

[@gptech2444]

Hi, just wanted to say thanks, was easy to folllow. I used your guide to add mergerfs to Proxmox. I had to use this line "pvesm add dir storage --is_mountpoint 1 --path /mnt/storage --content backup --shared 0" to get the drive to show up in proxmox, and also use apt-get command to install mergerfs.

[@distrodomain]

Oh nice what are you using the mergerfs mount point for, I've never set it up on proxmox, thank you for watching!

[@gptech2444]

@@distrodomain It's an old HP Z820 that I'm using as a home server. I just wanted to see if it would work with 3 drives for OMV. Not sure if this is best practice but I couldn't find anything to say it wouldn't work but I also couldn't find anyone that had tried it.

[@distrodomain]

I believe there ia a plugin for OMV for mergerfs and snapraid, I wouldn't recommend using the mergerfs mount point to store vm os drives, but as data drives for vm yes.

[@gptech2444]

@@distrodomain Yes, I'm using as data storage. I have a 1tb drive for the VM's. I'm still trying to get my head around what is best practice. I would've thought having mergerfs on the node would be best for use across different VM's.

[@balinttoth9287]

Excellent guide, thank you! Do you happen to have any recommendations for accessing the pool (/mnt/storage in your case) as non root user? Currently no files and directories are visible inside the pool when I log in as a non root user..

[@distrodomain]

I depends how are you sharing the mount point with your clients, I install samba and create a share, sharing the root mount point of mergerfs, you can control permissions and access with the smb.conf file, you can control the umask, and if it's broswable and public etc, I might make a video on how to do this exactly. Thank You for watching!

[@balinttoth9287]

For those of you stuck at the same problem, the non root user must be able to traverse the entire file path to the SOURCE directory as well, not just the merged directory. I have no idea why this is but after a few agonizing days, this was my only solution and now, everything work as expected. So for example if you merge '/some/folder*' into '/folders', and want to access '/folders' with a non-root user (that has read-execute privileges on it), you also need the non-root user to have read-execute privileges on '/some' !!!

[@distrodomain]

@@balinttoth9287 What I did is I added guest ok = yes force user = root It's less secure but this makes it so that any non-root user can traverse all paths, another thing you can do is use create mask = 0644 directory mask = 0755 or somthing similar in your samba config, and add your user to a group and make sure all files are created with this group, and adjust the permissiono of the files you already have. let me know if that helps :)

[@SAMUELDIY]

Thank you sir ,your video gave me much progress✔

[@distrodomain]

Glad to hear that :)

[@panther_mughil]

Thanks this was so helpful, Great Content, You earned a sub

[@distrodomain]

Glad it helped!

[@MIKA-rz9fq]

Damn, super helpful. Thanks man!

[@MobiusCoin]

Is adding more drives as simple as installing the drive, and adding it to the fstab config?

[@distrodomain]

Yes exactly! thank you for watching

[@Lafiro]

I appreciate the video. I followed the steps and I then ran a cp command from my external drive to my new mergerfs setup and somehow all the files only copies to a single drive, leaving another 3 empty. What went wrong? How can this be fixed? Manually moving files from one disk to another at this point?

[@distrodomain]

make sure you have category.create=mfs as a mount option on your mergerfs mountpoint, this is the option that tells mergerfs to put files on the drive that has more free space, you can check all the available policies on here github.com/trapexit/mergerfs, thank you for watching!

[@Lafiro]

@@distrodomain You know what, I have to double check the policy and then figure out how to resolve this. Again, wonder if I should just manually move files now from one disk to another or not

[@Lafiro]

@@distrodomain I had to make the change to add the line for most free space, but unfortunately, even after remounting everything is still being written to only one drive. How do I fix this?

[@distrodomain]

@@Lafiro paste here you fstab so i can take a look

[@Lafiro]

@@distrodomain /mnt/disk* /mnt/storage fuse.mergerfs defaults,nonempty,allow_other,use_ino,category.create=mfs,cache.files=off,moveonenospc=true,dropcacheonclose=true,minfreespace=200G,fsname=mergerfs 0 0 Just know that the category.create=mfs was not there when I first mounted this. I already copied files but never noticed they only went to one drive. I added that, did mount -a and tried adding a few files and they all still went to the same drive. First time doing this and I am still farly new to linux so I am lost now. I appreciate your help! Edit: I changed what you mentioned; even tried copying and pasting yours and now I can no lonher mount mergerfs. Just tells me its not found in fstab. I went back to my original line and same error. I no longer know what went wrong or how to fix this.

[@tamask]

Nice video. But I’m checking the docs, and allow_other and use_ino are deprecated or removed… Also, I don’t quite get what ignorepponrename does exactly. Did you run into issues by leaving it on default, which is false?

[@distrodomain]

Yes I noticed allow_other and use_ino are now enabled by default and no longer need to be added, what ignorepponrename basically does, it keep the file in the drive that is stored in even if you rename the file, it doesn't trigger your create policy, if you are using a non-path preserving policy. thank you for watching

[@tamask]

@@distrodomain Regarding ignorepponrename: I’m not sure it’s about triggering the create policy and causing the files to move due to a non-path preserving policy, when they shouldn’t be moved. I think, the issue is something else. The problem is when the destination path is on a different filesystem. In this situation, exactly a path-preserving policy is the one that causes the problem, because the target directory cannot be moved to the same drive as the source. That’s why the documentation suggests using a non-path preserving policy when there are problems with rename. The other thing I noticed is it says “With the msp or most shared path policies they are defined as path preserving _for the purpose of controlling link and rename's behaviors_ since ignorepponrename is available to disable that behavior.” This makes me think that ignorepponrename doesn’t even come into effect with non-msp or non-most shared path policies (and neither with any path-preserving policies). It’s only relevant for msp policies. The non-msp and non-path reserving policies would always just copy the target directory over, which is never a problem. And the path preserving policies would never do that, which is _always_ a problem. But maybe I’m getting it wrong, I’m not sure. I’ve read it like 15 times, and it’s still confusing. If it’s just a single file rename, where the directory path stays the same, I haven’t found any mentions that mergerfs would move the files between drives ever, regardless of what the create policy is. But again, I'm not sure.

[@distrodomain]

I'm not 100% sure either but I'm curious now, I will do some testing in my lab, and see what behavior I found

[@PasTrik]

thanks from france !!!

[@distrodomain]

You are welcome, thank you for watching :)

[@tristanbautista9705]

Hey man! I followed your tutorial and it worked but when I rebooted my server the mergerfs pool basically unmounts, the hdd is still detected and partitioned, but when I use df -h the mergerfs filesystem mounted on /mnt/storage is gone

[@distrodomain]

I would say check your fstab make sure the mergerfs is the last mount point in the list, and that the drives are named the same and matches the patern used for the mergerfs mount point, what does your fstab look like, thank you for watching!

[@sidewind131258]

I am curious, can I have multiple existences of merger disk running at the same time? Say I have an array? of 8x3 tb sas drives, of which 2 will be configured as parity Then I have 5x2 tb disks where I use 1? or 2? as parity Lastly I have 4x4 tb drives and 1 8 tb drive for parity (I actually do have that amount of disks available) Do I run that configuration as 3 separate arrays with each having its own merger/snapraid conf, or is it better to have them in a combined .conf? The reason I am asking, is that I have my win machine running m.2 disks and my Unraid machine on SSD disks, and I am having so much data that I want to use the 3-2-1 rule of backup, and have a friend hosting one of my backup machines in an location that is 350 km / 200+ miles away

[@distrodomain]

I would recommend having just 1 mergerfs mount point and using 2 drives for parity, I also recommend for your 2 parity drives to be the largest of the pool, so in your case if I read correctly you have 17 drives total and 4tb is the bigest size in your pool, so 2 4tb for parity the rest for data, you also might consider 3 parity drives to protect 17 drives. that is waht is recommended in here www.snapraid.it/faq#howmanypar, thank you for watching !

[@sidewind131258]

@@distrodomain first of all, thank you for your quick answer It is actually 18 disks 😁(1+4+8+5) Now to get this old reptilian brain wrapped around this: Use the 8tb plus 2 4tb drives as parity (sounds resonably as we are talking of 56tb total space) Now I will have something like this: disk 1+2+3 = parity disk 4-18 as data drive 1-15 /dev/disk/by-id/diskID /mnt/parity1, parity2, parity3 filesystem defaults 0 0 then /dev/disk/by-id/diskID /mnt/disk1->15, filesystem defaults 0 0 /mnt/disk* /mnt/storage fuse.mergerfs or use the mergerfs string you have provided Now hopefully the cabinet and ram comes monday and I can start playing 🥰🥰 Hups, one last question: Ubuntu server or what you are usin, I can't right now remember the distro name

[@distrodomain]

Yes that looks good, just make sure you don't mount the parity drives in your mergerfs mountpoint, I'm using rocky linux 9, it's a redhat equivalent, ubuntu should work too.

[@sidewind131258]

@@distrodomain why thank you my good man for putting up with a computer illiterate (well I DID build my first computer with resistors etc and a Zilog Z80 processor screaming away at 2.5 MHz) I might get back to you from time to time

[@ajfriesen]

You actually do not need to create partitions. You can just create a filesystem on the block device. Less complexity.

[@distrodomain]

Yes! you can create the filesystem just on the block device that woks too

[@trapexit]

This can be useful in cases where you may move the drive between a SATA to a USB-SATA adapter for example and the LBA size can change.

[@ajfriesen]

@@trapexit Oh this is interesting! Can you elaborate on this one? Is it better to have no partitions when you switch the physical interface? Did not hear about this case yet.

[@trapexit]

@@ajfriesen A lot of retail drives are still using a LBA size of 512 bytes when connected via SATA but it is common for some USB-SATA adapters to use 4k. Since partition offsets are set in blocks rather than bytes it can cause issues if jumping between interconnects which use different sizes.

[@AlexKretzschmar]

@@trapexit that is neat!

[@kang7014]

is it available in GUI?

[@distrodomain]

At the moment this tools is only available in CLI mode. Thanks for watching :)

[@mdd1963]

a file system with no redundancy or resiliency? I'll pass.

[@distrodomain]

Check out this video kzbin.info/www/bejne/h3eqdn6eZqZsZqs I made on snapraid, usually snapraid is used in combination with mergerfs to add the redundancy to the file system, mergerfs is your pooling solution and snapraid is your raid/redundancy solution. Thank You for watching.

[@jrmontz4363]

Greetings ! Which Distro are you using ?

[@distrodomain]

I’m using Rocky Linux 9, any redhat based distro is similar, thank you for watching

[@jrmontz4363]

@@distrodomain Thank You so much ! Your videos helped me A LOT !