4k format drives and zfs

P

pillai_hfx

Hi,

I have a new server getting delivered in the next few days with some 3TB Seagate drives. I am planning a raidz3 with this, which would get expanded with more raidz3 vdev's later. Once I follow the gnop option to create the initial raidz3 with an ashift of 12, is there a way to make sure that all future raidz3's also get an ashift of 12 automatically? Or is the ashift set for the whole pool? I checked the zdb config from one of my other freebsd FreeBSD zfs boxes with 5 raidz2's and it shows an ashift of 9 for each vdev. Using gnop and exporting and reimporting the pool is ok initially with no data in the pool, but I wouldn't want that with a nearly full and heavily used pool and at the time of expansion. So is there a way to avoid this procedure for later raidz3 vdev additions and make that ashift of 12 persistent for the whole pool?

One other question is when a drive fails and it gets replaced, is there anything special that needs to be done to add to the pool with an ashift of 12? (like using gnop again for the replacement drive?) It looks like it would be a while before 4k only drives come out. If it ends up being way too much trouble, I may not fiddle around with ashift, as the 512B emulation may not be going away soon. I can always create a new pool with 4k only drives later in the worst case instead of adding to the current pool.

Thanks.
 
update

After reading this thread (http://comments.gmane.org/gmane.os.solaris.opensolaris.zfs/49253), I am hopeful that the 7200.14 series drives might present itself as 4k sector drives. So hopefully I will get the ashift of 12 automatically. With an ashift of 9, future replacement for the failed drive in the vdev becomes a problem if only 4k sector drives are present in the market.

This is from that thread "Based on the numbers stamped on drive and Seagate support, yes the
7200.14 present 4k sectors and the 7200.12 have a jumper that switches between 512 and 4k; though I don't know if that means the disk is 4k or 512 internally."
 
Now that time has passed and lessons learned, for the sake of being accurate and helping future people who see this thread:

1) No, unfortunately, the Seagate 7200.14 series 3TB drives (at least the ones I received and used so far) present as 512b and ashift=9 by default. This probably applies to all of them, but the exact ST3000DM001 Part Numbers (since there are many) that I have seen this on are: 1CH166-571, 9YN166-500, 1CH166-301.

2) According to this Seagate Barracuda Data Sheet, bytes per sector = 4096, so I'd assume that's internally.
 
I just treated all of the new 'greenish' drives as 4k - created GPT partitions and used gnop accordingly when building the zpool.

Am getting good performance on a 10 x 2TB raidz2. Using Samsung, Seagate, Hitachi & WD drives.
 
You must log in or register to reply here.
Back
Top