Hi guys running freenas 7 on Microserver with 5 x 3TB HDD running ZFS Single Parity Raid 5.

I made this like months and months ago and have realised I forgot to tick the box 4k sector, how do i do this now without loosing my stuff as I have over 3TB files already on it?

you can't without destroying the raid volume.

lol @ freenas.

that was not what i wanted to hear :(

Just put your stuff into the cloud, change it, then let the cloud rain down upon you*.

*This WILL work, but there may be more time efficient methods of temporary storing the data you want to keep.

Did you really just suggest he upload 3Tb into the cloud mental? :)

Ahh yes, my bad it was Eorl not Trotters Independent Trading that asked about using the cloud.

Hi guys running freenas 7 on Microserver with 5 x 3TB HDD running ZFS Single Parity Raid 5.

I made this like months and months ago and have realised I forgot to tick the box 4k sector, how do i do this now without loosing my stuff as I have over 3TB files already on it?

how many months and months ago? are you sure the drives are 4k advanced format? cos google says freenas should use 4k sectors automatically if the drives support it

Yeah what Jim said - FreeNAS faq says that the checkbox is optional if all drives support 4K sectors.
I don't think there are any 3TB drives out there that aren't advanced format are there?

There's a couple of shell commands that you can use to check for sure anyway:
http://doc.freenas.org/index.php/FAQs

they are Hitachi 3tb5400rpm

Read up on a shift-12

Hitachi should be native 512k sector drives

for noobs like me, what does having 4k sectors do?

4K sectors handle large files much more efficiently but not so much small files, so you would probably want to use smaller sectors if you were say running a database. From memory with the last array I built I also needed to go 4k sectors to allow me to create a gpt ntfs volume larger than 16tb (i think).

In addition to improving storage efficiency for larger files, it's also to beef up error correction. Smaller disks with 512k sectors don't have enough physical error correction information when you move to the higher magnetic storage densities required for disks with larger capacity.

mm i not sure but think that s what the problem is?? when copying data across to the drive im only getting from 10-30mb sec

well did you check the freenas docs for how to check? daz linked it

also, when I googled this yesterday I saw a lot of zfs/samba tuning parameters being mentioned for improving file copying performance, many depending on the amount of ram you have

Tit are you running 4k A-shift12 sector alignment on your zfs pool? I can't remember if v15 of zfs supports it

How much ram do you have 2/4GB by chance? The rule of thumb is 1gb of ram for every 1tb of disk space in an ideal world. Are you running deduplication if so then you will want min 16GB

What are the specs of the machine?

When did you last scrub and how long did it take?

Have you check the seq read / write of the array itself with something like bonnie++?

Do you have plans to attach zil or l2arc to your pool?

http://groups.google.com/a/zfsonlinux.org/group/zfs-devel/browse_thread/thread/671325cc79dade71

http://forums.freenas.org/archive/index.php/t-66.html

I found some nics are f***ing s*** awful under freenas plus I find its problematic for the work I am throwing at it. I have since dropped freenas and opted for Solaris Express 11 performance nearly tripled.

Lastly are you running WD's greens if so then stop right now pick all of them up and throw them in the bin where they belong

Hey HerbalLizard.

I have Hatachi 3TB 5400rpm drives. I am running on HP MIcroserver with 4gb ECC ram, if i need to can upgrade to 8gb ram but at the moment its not even using 50%. I am running basic ZFS with Raid 5 1 drive partity. i dont know much about zfs but all i did is create the virtual disk then created pool that its, didnt tick any boxes or added different options.

when i run this command
zpool set cachefile=/data/zfs/zpool.cache NAS (change NAS to your pool name)
zdb -U /data/zfs/zpool.cache | grep ashift
i get error saying to something too many... (i cant remember exactly at the moment)

Im a big noob to linux first os i've used

post back when you remember, cos that information is useless as it is

The rule of thumb is 1gb of ram for every 1tb of disk space in an ideal world.

What on earth? Why?

The rule of thumb is 1gb of ram for every 1tb of disk space in an ideal world.

I thought that was only if you want to do ZFS de-duplication? Or is that true of ZFS in general?

Mostly dedup trog

And that rule of thumb came out of the freenas docs http://doc.freenas.org/index.php/Hardware_Requirements

Btw the 5k3000 drives are 512k

Have you tried another nic?

What's the client win7 I assume? What sort of switch which would lead to the next question does it support jumbo frames? I can tell you now that the broadcom nic in the mircosevers does not support anything over 1500mtu. I spent ages reading up on the nc107i and its supposed to support it but doesn't or at least the implementation in the micro server

Instead of SMB have you tried NFS/CIFS/AFP on another client machine. Maybe try downloading a live distro and testing nfs from there

https://help.ubuntu.com/community/SettingUpNFSHowTo#NFS_Client

What other processes are running on the box? Do you have appache spawning millions of child processes and eating ram like a wellfare kid in a candy store? As an example

Also that 5th drive is running off the sata port, is it running the modded ahci firmware on all drives or ahci on the ones on the sas controller and ide on the sata port. Or IDE over the whole lot cause that might be your bottle neck right there



last edited by HerbalLizard at 23:54:25 30/Jan/12

Thanks Herbal!

Yeah tried it on win 7 client using SMB and FTP. Have tried running NFS/CIFS/AFP.

I am running the following other programs on the USB stick, SABNZBD, Sickbeard, CouchPotato and Headphones. The speeds i get even when i try and shut down these applications. so maybe its the network card?

Considered it might be the drives? I've never had much luck performance wise with 5400rpm drives in raid, and I've known others in the same boat. The 3tb 7200rpm drives I've got in raid5 run like rockets.

I've got a microserver with the BCM5723 nic and have no issues near-saturating my gbit network to/from it, for what it's worth. it runs linux software raid, tgtd iscsi and samba

mmm maybe it is drives, i did pay $115 for 5 of them so they were pretty good deal :)

Have you tried installing bonnie++ so you can benchmark your arrays. I think Teq might be rolling with the same drives and I think his array was doing over 250mb read/write

http://www.coker.com.au/bonnie++/

Have you run a scrub on the pool yet? If its done a bucket load of resilvering the start testing the drives

Hitachi / IBM drive fitness tool will not work off the sas controller in the micro server so attach each disk one by one to a straight old sata port on another machine and see what happens

Sorry for the late reply but this is what i get herbal



freenas:~# zpool set cachefile=/data/zfs/zpool.cache NAS zdb -U /data/zfs/zpool.cache | grep ashift
too many pool names
usage:
set

the following properties are supported:

PROPERTY EDIT VALUES

available NO
capacity NO
guid NO
health NO
size NO
used NO
altroot YES
autoreplace YES on | off
bootfs YES
cachefile YES | none
delegation YES on | off
failmode YES wait | continue | panic
listsnapshots YES on | off
version YES

if you want to have play with it Herbal i am happy to give you login etc :)

Might have to wait until I get actual functional internet then sure no worries

sweet really appreciate it! just PM me :)

Sorry to bump but I found this thread while searching google :O

I finally got around the sticking 2x 3tb wd green drives in my microserver with FreeNAS 8.

Are these speeds normal ...

Running a 400MB file write on drive Z: 5 times...
Iteration 1: 57.43 MB/sec
Iteration 2: 51.12 MB/sec
Iteration 3: 57.84 MB/sec
Iteration 4: 53.3 MB/sec
Iteration 5: 49.6 MB/sec
------------------------------
Average (W): 53.86 MB/sec
------------------------------
Running a 400MB file read on drive Z: 5 times...
Iteration 1: 77.71 MB/sec
Iteration 2: 81.22 MB/sec
Iteration 3: 81.35 MB/sec
Iteration 4: 80.24 MB/sec
Iteration 5: 83.17 MB/sec
------------------------------
Average (R): 80.74 MB/sec
------------------------------

Should I just be content with it?

I'm running the drives in mirror with zfs but did not select 4k sectors (since I had nfi what it meant). Also have 8gb ram.

N.B - I am a huge noob at this stuff.

Yeah those speeds don't look too bad for green drives.

Should you be content with it? That's subjective and depends on your usage pattern.
I wouldn't expect a huge performance boost without spending money on non-green drives.
If you're not transferring a metric s***load of data on and off those drives every single day then I don't think it's worth worrying over.

Thanks Dazhel.

Would enabling 4K sectors speed things up? I realise I'd have to start over with my volume so I need to know if its worth it.

Thanks Dazhel.

Would enabling 4K sectors speed things up? I realise I'd have to start over with my volume so I need to know if its worth it.

Probably give you slightly better performance may not be noticeable though. It depends on what your'e going to use your NAS for, if you are going to mostly store large files i.e. videos, audios, pictures then it'd be worth it as it'll be more efficient to use large block size. But if it's going to be for things like word documents etc then standard block size is better.

Yeah I'm using to stream all my media to XBMC.

I'll give it a go. Thanks guys.

well as was already pointed out in this thread, unless you explicitly uncheck the 4k sector option during setup, it should already be enabled if your drives support it

Missed that one :S In that case I'll just leave it alone. Thanks again.

Should I just be content with it?

You've got me curious to get off my ass and benchmark my array home, might take a look after work if I remember.

I'm looking at upgrading to Freenas 8 when 8.2 when it comes out

Just tried running some read / write benchmarks on my array, not too bad.

WindowsDeveloperPreview-64bit-English-Developer.iso
Size: 5197092864

Read:
Transfer Rate: 908.086 MB/s
Transfer Rate: 916.482 MB/s
Transfer Rate: 1197.761 MB/s
Transfer Rate: 909.085 MB/s
Transfer Rate: 929.719 MB/s
----------------------------
Transfer Rate: 972.226 MB/S
----------------------------

Write:
Transfer Rate: 565.985 MB/s
Transfer Rate: 565.018 MB/s
Transfer Rate: 566.373 MB/s
Transfer Rate: 566.115 MB/s
Transfer Rate: 565.146 MB/s
----------------------------
Transfer Rate: 565.727 MB/S
----------------------------

http://andysworld.org.uk/2011/08/25/oracle-solaris-express-11-versus-freenas-quick-cifs-performance-tests/

The 3tb 7200rpm drives I've got in raid5 run like rockets.

Bloody hell, tspec, you don't say! how many spindles is that?