[zfs-discuss] L2ARC and SLOG on HW RAID with writeback cache
Edward Ned Harvey (zfsonlinux)
zfsonlinux at nedharvey.com
Fri Apr 27 17:42:06 EDT 2018
> From: zfs-discuss <zfs-discuss-bounces at list.zfsonlinux.org> On Behalf Of
> Gionatan Danti via zfs-discuss
> given a system with a proper, writeback-enabled RAID controller which
> can both create a classical array on some disks *and* passthrough some
> other disks, is it a good idea to place L2ARC and SLOG on the HW array?
The only bad performance configuration is to have a pool made of HDD's, with no slog, and write-through. If any one of those things changes - your pool is made of SSD's, or you have a SSD slog, or you enable write-back, then you'll have pretty good performance. That being said, having an SSD slog is better than having write-back, due to the way zfs does write aggregation. If possible, your best choice will be to use SSD slog, and *disable* write-back on all the HDD and SSD's. That is, you should use write-through, and SSD slog. Zfs benefits from, and performs better, by knowing for real, when its data is on disk as opposed to stored in hba write-back memory.
You can certainly manage root on zfs, but if it's easier for you to run hw raid on your boot disk, that's reasonable. You won't have some of the benefits of integrity checking and snapshots and so on, but it's your choice. Nobody's going to force you to use zfs. ;-)
More information about the zfs-discuss