I plan to have one master server including several HDDs (for RAID5, probably all homogeneous with 4TB each), and one remote slave server containing only one big capacity HDD, or a bunch of disks not necessarily all the same capacity. These 2 servers would have encrypted disks. It would be for running web/mail/media/git services etc, hosting software ISOs and video files larger than 4GB, and I'd have to set different rights/quotas for my family members. Hot swap/addition of a disk would be appreciated (don't know if this rather lies into OS choice though)

(I remember having corrupted an external drive and all its data, probably either or both because conflicts when alternating plugs into linux ext4 and windows ntfs machines, and letting it plugged when turning machines on sleep mode/hibernation.)

1) Assuming common filesystem for windows 10 is NTFS and for debian/xubuntu/arch 2020 is ext4 (idk what it is for macOS iOS and Android), what would be the optimal file system for the servers please, if any? If there isn't such, in what filesystems should I reinstall my different client machines for compatibility at least between each of them and the server (at most between any combination of 2 of them)?
2) Do the softwares pushing incremental backups that are installed on the clients have to be considered for this choice? Or is it usually only the servers which pulls everything anywhere?
3) Raid5 implies one disk crash tolerance, but are there actually ways to be notified of such crash ? Otherwise I'd rather choose other RAIDS with more tolerance if I don't have access to the server for a while.


Use ZFS instead of a parity RAID (many reasons including performance, copy-on-write, snapshots and extensive data integrity), enable virtualization on host (KVM, Hyper-V role, bhyve or whatever you're comformtable dealing with), isolate all the operating systems inside their own VMs and share content between your VMs using NFSv4/SMB3 protocols with either some dedicated file server VM or with some file server services running directly on your host. The only file system all of the mentioned operating systems can read/write is FAT and... it's a joke in 2020!

| improve this answer | |
  • thank you so much for the answer, I guess it's the right one but I allowed myself not to approve it the time I document myself, for welcoming other ideas. Few questions so far from the few conferences I looked at : why do people add storage to the pool by pair of disks (= adding 1 mirror each time) instead of a single disk in a RAID ? Do we agree that we don't have to pre allocate size of partitions and that they dynamically expand in zfs? Are there ressources for beginners as I don't understand most of the technical terms in the video conferences/books such as "checksums", "deduplication",..? – MKSwonderer Jun 4 at 13:37
  • How do you expect to align f.e. 256KB sized RAID strip with say 7 disks in a RAID group? – BaronSamedi1958 Jun 4 at 13:42
  • would a RAID (classical meaning, e.g RAID5 with 5TB available for 6x1TB disks and 1 failure handling, instead of a RAIDZ1 with 3TB available because of 3x pair/mirror of 2x1TB disks and 1 failure handling) of exFAT-formatted disks be able to donload/upload friles from all the win/linux/macOS/android/iOS clients? – MKSwonderer Jun 4 at 13:43
  • Sorry i really know nothing at all on both drive hardware and RAID so can't answer but i'd be glad if you could route me to some useful ressources – MKSwonderer Jun 4 at 13:47
  • You can configure RAIDZ1 to handle the same capacity as your hardware RAID5. While you might strip drives for performance you don't need to. Keyword "stripped vdevs". Read more here: zfsbuild.com/2010/05/26/zfs-raid-levels – BaronSamedi1958 Jun 4 at 14:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.