Finetuning ext4 on Raid5

JPT · 21. November 2019

Hi,

until now I simply had Debian on my NAS but now I decided to build a RAID and install OMV.

So, OMV does not use LVM on top of RAID, is that correct?

May I edit the OMV section in /etc/fstab?

this explains tuning the ext4 fs when it's sitting on a Raid.
So I calculated from this thread:

Code

mkfs.ext4 -b 4096 -E stride=128,stripe-width=256 -O dir_index -L RAID /dev/md0

Now to the mount options.
OMV uses

Code

/dev/disk/by-label/RAID /srv/dev-disk-by-label-RAID ext4 defaults,nofail,user_xattr,noexec,usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0,acl

You are aware that the short form of this would be to use LABEL=RAID?

nofail and noexec are fine.
three quota parameters - I should not touch them, as they for sure interact with OMV quota somewhere
acl - this is access permissions that are more flexibel than chmod? never dealt with them. won't touch.

thanks for any help
The thread linked above recommends mount options:

Code

data=writeback,noatime,nouser_xattr

This does the following

data=writeback - disables journaling for the data, the file content. Only metadata, ie the filesystem changes are journaled.
noatime - does not touch the access time ever. relatime only touches the access time if it's older than the mtime I think, so once for every file. (I use relatime on my SSDs)
nouser_xattr - extended attributes. no idea what they are good for!

For a home NAS I like this approach for reducing write access...
Anything that conflicts with OMV? any other comments?

ah, another question. the thread I am always referring to...
it talks about the fs options stride and stripe-width. But all kernel doc I found only talks about stripe. how come? The effective mount options now show stripe=256.

ryecoaaron · 21. November 2019

Zitat von JPT

So, OMV does not use LVM on top of RAID, is that correct?

You can but why would you need to? OMV doesn't use partitions with raid. So, if you expand the array, you can just expand the filesystem. LVM would just be another layer.

Zitat von JPT

May I edit the OMV section in /etc/fstab?

You can but it will be reverted.

Zitat von JPT

this explains tuning the ext4 fs when it's sitting on a Raid.

As long as you mount the filesystem in the OMV web interface, it is fine to create the ext4 filesystem from the command with your customizations.

Zitat von JPT

You are aware that the short form of this would be to use LABEL=RAID?

I'm sure Volker (author of OMV) is aware but what would that help?

Zitat von JPT

this is access permissions that are more flexibel than chmod? never dealt with them. won't touch

I hate ACLs. Good idea

Zitat von JPT

The thread linked above recommends mount options:

Edit the entry in the mntent section of /etc/openmediavault/config.xml and the run omv-mkconf fstab (on omv 4.x) or omv-salt deploy run fstab (on omv 5.x).

Zitat von JPT

ah, another question. the thread I am always referring to...
it talks about the fs options stride and stripe-width. But all kernel doc I found only talks about stripe. how come? The effective mount options now show stripe=256

I think you should just try the default options. It will easily saturate gigabit and is good performance for 99.9% of users.

Jetzt mitmachen!