Monday, September 28, 2009

Thumper features in EON ZFS storage

With binary kit 122, the package SUNWhd (tools and driver download) used in thumper was added to EON. With this package you get the capability to tell a great deal of metrics about your disks. It gives the ability to control read and write cache. It allows to tell firmware revision, vendor, serial and very importantly, the temperature of the disks.

We know that higher temperature disks, perform less efficiently. In my DIY ZFS storage case, hdadm showed that there was insufficient cooling or air was not being moved efficiently through the cage as the temperature of the 15K rpm drives continuously inched towards the clip or shutdown temperature. The clip temperature can be seen in the output of:
smartctl -d scsi -a /dev/rdsk/cXtXdXsX
... excerpt
Device supports SMART and is Enabled
Temperature Warning Enabled
SMART Health Status: OK

Current Drive Temperature: 48 C
Drive Trip Temperature: 65 C
Manufactured in week 31 of year 2006
Re-orientation of the drives and re-positioning a 20MM fan seems to be holding at 53 degree C. I'll focus on getting temperatures down to the 36-40 degree C range. For more use see hdadm which is a wrapper script for calling the hd binary.
hdadm display

Device Serial Vendor Model Rev Temperature
------ ------ ------ ----- ---- -----------
c0t1d0p0 DNL3P68018NM FUJITSU MAX3036NC 5D03 53 C (127 F)
c0t2d0p0 DNL3P6400SDR FUJITSU MAX3036NC 5D03 58 C (136 F)
c0t3d0p0 DNL3P68018NN FUJITSU MAX3036NC 5D03 54 C (129 F)
c1t1d0p0 DNL3P6400SEE FUJITSU MAX3036NC 5D03 49 C (120 F)
c1t2d0p0 DNL3P68018BC FUJITSU MAX3036NC 5D03 51 C (123 F)
c1t3d0p0 DNL3P68018NV FUJITSU MAX3036NC 5D03 46 C (114 F)
c2d0p0 HNCF128MMG=STCB21M84025C03356C1 MG 84025C03356C1 - None
A hard drive performance test can be run using:
hdadm diskqual

3 comments:

Unknown said...

58 C for a harddisk is a bit to hot for my liking!

Andre Lue said...

Ramoonus,

True. More to the point is everything seemed fine until hdadm revealed that facts on how hot the drives were.

Showing that the drives should be made cooler. Great feature to have.

Garen said...

It's really handy to get this kind of info.

I can easily see using a script and cron job to periodically grab the temperature data and make some simple charts, and send myself an email if something gets out of whack.

I wonder why the hd utils aren't included in the standard OSOL distro?

Super useful.