Filesystem Recommendations - Best Practices

    • Filesystem Recommendations - Best Practices


      I am a newcomer to the OMV forums. I have been lurking and testing OMV for a period of several weeks now with great success. I evaluated FreeNAS, OpenIndiana, NAS4Free, OMV and piecing together my own solution. I began with a VMware instance on my laptop and quickly realized the systems potential. From there I repurposed a physical server into OMV as it seemed to be the most capable and simplest of them all.

      Current configuration:
      Intel Core i5 2500K 3.3 GHz
      8GB RAM
      1000Mbps wired network
      3x 2TB HDs (retiring to another system)
      ESXi 5 hypervisor for management

      Services / Plugins:
      iTunes/DAAP (forked-daapd memory usage needs to be addressed, it hangs on text files causing massive CPU load time and memory draws)
      Apple Filing
      Local DNS / DHCP

      My testing is concluding and I plan to dedicate this system for the services above. But before I do, I am contemplating filesystems and resiliency. My current disks are using a mix of XFS and ext3 and the performance is quite good. Being a Linux user for many years I have typically used the ext filesystem and software RAID5 with success.

      Today I purchased 3x 3TB Western Digital Red drives to put into this system and prior to deploying them into "production" I wanted to ping the more experienced NAS group for guidance. My original intention was to create a software RAID5 array and layer in LVM with ext4 or XFS, all through the OMV GUI. Obviously I rather do this once and "set it and forget it" and merely grow the array and LVM with additional Red 3TB drives as I require more space. I anticipate growth to at least 15+TB and want to architect the solution to withstand up to 2 physical drive failures and maintain the ability to easily grow the filesystem as drives are added.

      I invite information, recommendations, guidance from the group as I sit anxious and ready to get this thing online.

      Most of all, a big thanks to the developers, contributors, and support of this community. I am very impressed with the software capabilities, usability and quality compared to the commercial variants available. :applause:

    • Re: Filesystem Recommendations - Best Practices

      okay some short answers from my side.

      I would opt to ext4 allways now. You can tune it to the underlying raid level which should benefit you.

      If you want to survive two disk failures, you need to go RAID6. For now I am not sure if you can use RAID5 as a start and then convert it to raid 6. Should work but not sure.

      If you want to read something about tuning, have a look at this thread:
      Everything is possible, sometimes it requires Google to find out how.
    • Re: Filesystem Recommendations - Best Practices

      Thanks for your response SerErris. And thank you for the tuning link, this information is very valuable. However, it seems to have been moved here:

      I understand RAID6 is the best option for my 2 disk failure requirement (apparently until 2019 fwiw). I have begun investigating and studying the conversion process from RAID5 --> RAID6 and although it seems straightforward and supported by mdadm I am eager to hear the experiences of others, especially those on this forum. With my 3 HD setup I will initially build a RAID5 array and migrate to RAID6 later this year if deemed possible and not too much of a risk to the data.

      I've gone ahead and created my disk array, LVM and FS and am testing performance after using your tuning guide.
    • Re: Filesystem Recommendations - Best Practices

      I have used the conversion many times with no issues. It works and is not very complicated to achive. It takes some time but it is not an issue.

      If you have more disks, you also may consider to use two raid5 instead of one raid 6 and create to PVs on it to use it in LVM. But up to 6 - 8 drives, Raid6 should not be an issue.

      Everything is possible, sometimes it requires Google to find out how.