MergerFS - Not writing files to newly added disk

  • Hello,


    I recently installed OMV 5.x on my NAS (from Ubuntu 18.04) utilizing an existing mdadm RAID1 config. One of these RAID disks was starting to fail (high amount of unrecoverable sectors). I decided to buy 3 x 3TB (Ironwolf) drives and configured two into a mergerfs pool (with a most free space create policy) and enabling SNAPRAID with a single 3TB parity drive. All went well with this configuration. I copied all my data off the RAID array onto the mergerfs pool without issue. I then decided to remove the now empty RAID array, take the one healthy disk (reformat) with a new file system and add to the mergerfs pool.


    Now the problem. When I write files to the pool, the new disk is not receiving any blocks. I keep on having to run the tool mergerfs.balance to resync the data. This works to within 2% of free space (default with the balance tool), but the problem still persists. I've rebooted the NAS thinking that a re-mount of the pool is required but this has not been effective either.


    Any suggestions on how to troubleshoot? The drive in question is DATA2. Right now there is 5% difference but as you can see with DATA3 and DATA4 they are closely matched.


    /dev/sda1 1.8T 507G 1.3T 28% /srv/dev-disk-by-label-DATA2

    /dev/sdc1 2.7T 893G 1.9T 33% /srv/dev-disk-by-label-DATA3

    /dev/sdg1 2.7T 892G 1.9T 33% /srv/dev-disk-by-label-DATA4

    /dev/sdf1 2.7T 811G 1.9T 30% /srv/dev-disk-by-label-PARITY1

    MEDIA_UNION:eb06833a-73ce-437e-87c7-d3c2c5c2ae8b 7.2T 2.3T 5.0T 32% /srv/eb06833a-73ce-437e-87c7-d3c2c5c2ae8b


    Thanks!

  • DATA2 has 1.3T available. DATA{3,4} have 1.9T. Why would you expect DATA2 to be chosen as the target for a create with policy `mfs`?


    I'm assuming that since DATA2 use% is lower, and thus has more free space on the disk from an overall % perspective, it would be the likely target. Based on your response, I'm not understanding how mfs works within what you designed. Is there a policy which creates a write policy based on %use?


    EDIT: Yep... I had the wrong policy. Should be LUS. Thank you for responding. Great product btw.

  • jayman81

    Hat das Label gelöst hinzugefügt.
  • I would be glad if it would be possible to select percentage-based policies (by now I see eppfrd, msppfrd and pfrd available).


    Thank you for the fantastic work on OMV and kind regards :)

  • I would be glad if it would be possible to select percentage-based policies (by now I see eppfrd, msppfrd and pfrd available).


    Thank you for the fantastic work on OMV and kind regards :)


    Have you tested any of these percentage-based policies yet? I generally have used MFS, but had been hoping for percentage-based writing for a while. I guess this is not exactly the same as writing to whichever disk has the lowest percentage used, since it still seems to choose the disk(s) to which to write based on gross free space.

Jetzt mitmachen!

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