Create a Filesystem with a blocksize greater than 4096

1. offizieller Beitrag

    Hello,

    I am working to create an NVME stripe in my system and have gotten to the point where I need to put the filesystem on the stripe that was created with mdadm. I would like to use a block size larger than 4096. Specifically I would like to use a blocksize of 16384.


    I created the file system with mkfs.ext4 in the command line and it appears in the GUI. However when I attempt to mount the FS in the GUI is returns an error and can not mount the FS.


    I have since deleted the filesystem and would like to know if there is a way to set the blocksize via the GUI or how to property create the FS in the CLI so that the GUI can then be used to mount it etc.


    The reason for the large blocksize is that everything I have read online about doing this says the block size of the FS should be LBA x N(number of disks). THe NVME drives have been reformatted to use 4096 LBA.


    System is OMV 5.6.24-1

    4 x 1TB NVMEs

    64GB Ram

    Intel Xeon 6130


    I have an existing Raid 6 for my shares and I am booting from a separate 1TB NVME in the system.


    I am still learning my way around Linux so I most certainly doing something wrong and any help or advice will be welcomed!

    Where did you read this? I would not do that, if you do not have very special requirements on the file system. Bit give me a ponter to understand you thoughts.

    If you got help in the forum and want to give something back to the project click here (omv) or here (scroll down) (plugins) and write up your solution for others.

    Zoki,

    I can not seem to find the page that mention that. One day I will learn not to have 50 pages open while researching. When I find it again I will post it.


    I do not have any special requirements just trying to get the stripe to perform better than a single NVME that is a member of it. I am going ahead and formatting the raid via the gui to get it so I can continue testing based upon the updated LBA.


    Since this is a stripe that will hold trash data I can start again if I need to change the block size.


    As to the question about the filesystem not being able to be mounted in the GUI is there something that can be done about that?

    • Offizieller Beitrag

    I did some searching regarding this, it seems this can be beneficial to improve I/O particularly if using databases or large file access but I would have thought this would be more applicable to mechanical drives.


    As to mounting the file system, if you have created it in the cli then it should be possible to mount the same way, but I think omv takes exception to this not sure.


    As you've stated your stripe will hold trash data I fail to see the benefit in what you're attempting to do.

    Raid is not a backup! Would you go skydiving without a parachute?


    OMV 6x amd64 running on an HP N54L Microserver

    Having large block sizes is detrimental to the number of inodes and wastes space, as every file is splitt unto parts of blockize, rounded up. So a File of 1 Byte lenght takes one block.


    A file of 16385 bytes takes two blocks of 16k bytes or 5 blocks of 4k.


    If you have mechanical drive, the block size is related to the physical geometry of the drive, but not the number of disks.

    If you got help in the forum and want to give something back to the project click here (omv) or here (scroll down) (plugins) and write up your solution for others.

    Zoki

    Thanks I reverted the LBA back to 512k and am going to see about getting the performance out of a single nvme drive and then move on from there.


    This has all been a bit of testing as I migrate to my new server and I am trying a few different things while I get it all setup.


    I follow the byte length information, and I am writing large files which why I wanted to see about increases the size. However my testing after setting up the 4096 LBA and creating the stripe with a 4096 block yield slower performance than the LBA at 512 and the blocksize of the FA at 4096. I will keep playing with it.


    Right now I am moving over my JBOD so that has taken most of my time so far today.


    Thank you again!

    Zitat von geaves

    As you've stated your stripe will hold trash data I fail to see the benefit in what you're attempting to do.

    Sorry I missed this question earlier. My use is very basic, recently upgraded to 5Gbps Internet and wanted to actually see that written to a system at something close to that speed. :D I know many sources on the net will be hard pressed to source data at this rate, but of course the answer to all this is because I can, maybe LOL.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!

Benutzerkonto erstellen Anmelden