zfs on two disks

Hi,

I will obtain a new computer for my office and I'm wondering which is the best configuration/filesystem for the two disks it has, namely:

300GB Western Digital Raptor 10.000 rpm sata II 16 Mb cache
1TB sata II Western Digital 7200 rpm 64 Mb cache

I have plans of use the 1 Tb disk to create a ZFS data pool that would be mounted in /usr/data with the following hierarchy:

/usr/data/current
/usr/data/{blah0..blah1}

Is it possible to have /usr/data/current mirrored in the 300GB disk with RAID1, together with the rest of the system? Is it possible/good idea to have both disks completely ZFS, including the boot and system for a FreeBSD 9 amd64?

Thanks in advance
 
You can set this up so that you've got a ZMIRROR covering just 300GB of the 1TB disk and the entire 300GB disk. I've done that in the past and it does work. ZFS itself doesn't know whether it's on a partition or a full disk. One of the gotchas is that you can end up in a situation where if you disconnect the 300GB drive and run just the half on the 1TB drive where you can't reattach the 300GB drive since it isn't quite large enough.

If you're mindful to make sure that the two are the same size it isn't much of a problem.

You definitely can have both disks be completely ZFS, but at this point it isn't supported in sysinstall and requires some tricks to pull off. I haven't personally had any trouble doing it. YMMV.

As far as whether or not it's a good idea, I'm not really qualified to answer that.
 
I don't understand why you would mirror together two such drastically mismatched disks. Put the 10000 RPM disk to some good performance-sensitive use, you're going to slow it down by mirroring it with a 7200 RPM disk. Or sell it to somebody who will, and get another 7200 RPM disk to mirror it with. What you're thinking about will give you the speed of your slow disk and the capacity of your small disk, worst of both worlds.
 
Could create 2 pools, one per drive. Keep in mind, ZFS supports a huge number of snapshots. If you simply want two copies of everything on the 300 gig drive, then there's a way to use "zfs send" and a cron task to implement a form of mirror. Some will complain that this isn't an offsite backup, but it is a good method for recovering "accidental deletes" or reverting to prior states of work.
 
Thanks guys, now I realize that is not my option. I will work with the 10000 rpm only, and use the 1Tb for backups (old data) periodically made by my own. Now, when I want to work with some old data sets I would like to move it to the 10000 rpm disk, How should I done this within the zfs framework? I am new to freeBSD and the only reason to use ZFS is to keep my data safer (also UFS is not good for disk with large cache right?). At the end How is the best way to send/copy my old data in the big tank? zpool send? making snapshots is an alternative?
Another question... If I would use one disk for keeping old backups, is it good idea to tune it for spindown after some time? Is it safer for a disk that I will access only four or five times per day? or is the spindown itself worst than having the disk at 7200 doing nothing?
 
Zhwazi said:
I don't understand why you would mirror together two such drastically mismatched disks. Put the 10000 RPM disk to some good performance-sensitive use, you're going to slow it down by mirroring it with a 7200 RPM disk. Or sell it to somebody who will, and get another 7200 RPM disk to mirror it with. What you're thinking about will give you the speed of your slow disk and the capacity of your small disk, worst of both worlds.
That's not true. You're right about the speed, but the capacity is a non-issue. You match the 300gb disk with a 300gb partition on the 1tb disk. Any loss of capacity is pretty minor. The only gotcha is that you could be in a situation where you can't add one or the other back if you have to detach the smaller partition from the mirror. I've done that in the past when I had a 1.5tb disk and a 1tb disk which I needed to mirror together. I wound up with a mirror of 1tb and an extra 500gb or so in unrelated filesystems.
 
I can't answer your question about spin down because I just don't know the answer. But, I've had a similar question in mind for a while.

Using "zfs send" and "zfs receive" has worked well for me in the past. I haven't used it extensively, but I would be confident in relying on it.
 
The thing is I don't know how data are safer (besides the power consumption), with disk spin down or running all time...
Concerning the zfs system, I have to read more about, thanks anyway ian-nai.
 
Spinning disks up and down all the time wears them out faster, you may save some energy but you'll have to replace the disks more often.
 
Back
Top