Hacker News new | past | comments | ask | show | jobs | submit login

"production" at home on Debian 11, previously on FreeBSD 10-13. The weirdest gotcha has been related to sending encrypted raw snapshots to remote machines[0],[1]. These have been the first instabilities I had with ZFS in roughly 15 years around the filesystem since switching to native encryption this year. Native encryption seems to be barely stable for production use; no actual data corruption but automatic synchronization (I use znapzend) was breaking frequently. Recent kernel updates fixed my problem although some of the bug reports are still open. I only moved on from FreeBSD because of more familiarity with Linux.

A slightly annoying property of snapshots and clones is the inability to fully re-root a tree of snapshots, e.g. permanently split a clone from its original source and allow first-class send/receive from that clone. The snapshot which originated the clone needs to stick around forever[2]. This prevents a typical virtual machine imagine process of keeping a base image up to date over time that VMs can be cloned from when desired and eventually removing the storage used by the original base image after e.g. several OS upgrades.

I don't have any big performance requirements and most file storage is throughput based on spinning disks which can easily saturate the gigabit network.

I also use ZFS on my laptop's SSD under Ubuntu with about 1GB/s performance and no shortage of IOPS and the ability to send snapshots off to the backup system which is pretty nice. Ubuntu is going backwards on support for ZFS and native encryption uses a hacky intermediate key under LUKS, but it works.

[0] https://github.com/openzfs/zfs/issues/12014 [1] https://github.com/openzfs/zfs/issues/12594 [2]https://serverfault.com/questions/265779/split-a-zfs-clone




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: