Comments on: Shenanigans with ZFS flushing and intelligent arrays…

By: Jason

Jason — Thu, 24 Sep 2009 23:45:26 +0000

On a 6140 I’m not sure. On an older StorageTek FLX280 ZFS was a lot faster. I believe both are xScale procs.

By: John

John — Sat, 12 Sep 2009 10:29:23 +0000

Hi There, we have a 6140 with expansion tray, 32×450GB 15000K drives attached to 2×5240’s 1.4GHz. I asked the project for a week of performance testing but was turned down. At the time I believed it made sense to allow the built in raid controllers to do the work and created 4xhardware-RAID6 LUNs and presented two luns to each T5240 in RAID0.

Each host runs 8 Oracle 10G databases and 3×10gAS app servers, all with probably different load requirements. DBA’s can’t tell me those requirements though. What I can tell you is the machines have plenty of cpu and approx 10GB real memory of head room with all apps running if I wanted the server cpu’s to do the raidz2 crunching…

Has anyone ever compared a 6140 with each disk presented as a LUN with raidz2 (mentioned above), to a 6140 with hardware raid6?

By: Jason

Jason — Thu, 15 Jan 2009 19:33:37 +0000

Hi Wang, thank you for catching that. Just moved the blog from Nucleus to WordPress a few weeks ago, and it seems some of the old URLs are broken. I’ve fixed them in this entry, so it should work now.

By: Wang

Wang — Thu, 15 Jan 2009 05:35:24 +0000

Thanks for the useful article!

In you article: “*** Please check out the update to this article [here]! There’s a better way now! ***”

But I can’t find the “better way” in that link, please point a way to me

Thank you very much!

By: Rejean

Rejean — Tue, 07 Oct 2008 17:03:11 +0000

Solaris 10 11/06 and Solaris Nevada (snv_52) Releases
Set dynamically:
echo zfs_nocacheflush/W0t1 | mdb -kw

Revert to default:
echo zfs_nocacheflush/W0t0 | mdb -kw

Set the following parameter in the /etc/system file:

set zfs:zfs_nocacheflush = 1

Risk: Some storage might revert to working like a JBOD disk when their battery is low, for instance. Disabling the caches can have adverse effects here. Check with your storage vendor.
[edit] Earlier Solaris Releases

Set the following parameter in the /etc/system file:

set zfs:zil_noflush = 1

Set dynamically:
echo zil_noflush/W0t1 | mdb -kw

Revert to default:
echo zil_noflush/W0t0 | mdb -kw

Risk: Some storage might revert to working like a JBOD disk when their battery is low, for instance. Disabling the caches can have adverse effects here. Check with your storage vendor.

the second method is working.
source:
http://www.solarisinternals...

By: Brett Morrow

Brett Morrow — Wed, 02 Jul 2008 08:52:55 +0000

With the infortrend units we saw worse performance with zfs. The problem lies in the cache flush and Infotrend confirmed it can not be turned off. I asked them to make a not about possible including this. We did get a lot better performance with adding this to /etc/system:

set noexec_user_stack = 1
set noexec_user_stack_log = 1
* NOTE: Cache flushing is commonly done as part of the ZIL operations.
* While disabling cache flushing can, at times, make sense, disabling the
*ZIL does not.
* If you tune this parameter, please reference this URL in shell
* script or in an /etc/system comment.
* http://www.solarisinternals...
set zfs:zfs_nocacheflush = 1

By: Raj

Raj — Sat, 03 May 2008 07:40:50 +0000

I have a disk array unit of 24 1T disks with 512MB cache and battery backup. The unit is from infortrend and I have created 1 big LUN of Raid6(h/w raid in infortrend) and attached to a solaris 10 machine and am using single LUN zfs pool. Doing this I am seeing very high IO performance compared with using ufs fs.

I am not using Raid-Z since OS only sees 1 big disk. Is it possible to get much better performance if I make 1 LUN per disks and share all the 24 disks in the unit as 24 LUNs and then make zfs pool of Raid-Z using these 24 LUNs?

By: tim

tim — Sun, 20 Apr 2008 15:57:29 +0000

>any tips on using Santricity with a Sun STK 6130: do you need specific firmware on 6130 to have SANtricity recognizing the this particular storage?

You can't use SANtricity with a 6130 because it is not an Engenio box. STORade or CAM only.

By: Andy Lubel

Andy Lubel — Wed, 05 Sep 2007 08:47:20 +0000

cant seem to use the IBM storage manager for managing the 6130, at least the package i saw, it wasnt generic and was looking for a ds4100.. can you tell us in more detail how to get this download from IBM?

By: Ulrich Gräf

Ulrich Gräf — Fri, 29 Jun 2007 13:02:04 +0000

zfs:zfs_nocacheflush is supported in Nevada and will probably be in the next S10 update. Unfortunately it swiches off the »flush write cache scsi command« for all disks in the system – which is a bad idea if you have also zpools on local disks with write caches, like the newer SAS or SATA disks. These commands help a lot if you have this mixture.