Date: Wed, 21 Jun 2017 07:45:02 -0600 (MDT) From: Warren Block <wblock@wonkity.com> To: Peter Ludikovsky <peter@ludikovsky.name> Cc: freebsd-questions@freebsd.org Subject: Re: New User, new server Message-ID: <alpine.BSF.2.21.1706210733220.60980@wonkity.com> In-Reply-To: <800e15b2-d7f5-d339-bd77-862e9d0cab5b@ludikovsky.name> References: <800e15b2-d7f5-d339-bd77-862e9d0cab5b@ludikovsky.name>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 20 Jun 2017, Peter Ludikovsky wrote: > 1) The new machine comes with a 128G SSD, in addition to the 2 4T HDDs > from the older server. I'd like to set up ZFS root, with a slice of the > SSD as ZIL and L2ARC, and the root mirrored across the SSD and the 2 > HDDs. Does this make sense, and if so what would be the ideal slice > layout? Or should I just use the whole SSD as ZIL/L2ARC? Don't mirror an SSD with hard drives. It will work, but cancels the benefit of the SSD. You don't say how much RAM the system has. Adding L2ARC without a decent amount of RAM is actually worse than nothing. ZIL is built in. An SSD to cache ZIL is called a SLOG. ZIL (and a SLOG) are for improving the speed of synchronous writes. Generally, that is rare unless you are using NFS and virtual machines. Also, a SLOG that is worthwhile usually needs a low-latency SSD that is used exclusively for that purpose. TLDR: unless you have at least 32GB or 64GB of RAM, adding L2ARC is questionable. A SLOG is usually only helpful if you have lots of synchronous writes, which is rare. Even then, it should be a fast, dedicated SSD just for that. http://doc.freenas.org/11/zfsprimer.html > 2) Moving data from the old machine. Can I run zfs send/receive to get > the ZFS on Linux datasets onto FreeBSD, or do I need to (r)sync? zfs send | zfs recv usually works. It depends on the feature flags on the originating system.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.21.1706210733220.60980>