[ltp] high disk temperature Issues in T23

[Paul Seelig]

Hi all,

this message is aimed mostly at sharing the solution i found for a too
high disk temperature with a ThinkPad T23. Sorry in advance for the
chatty prose style.

I am the proud owner of an old T23 (2647-9RG), which has been maxed out
to 1GB RAM and a 160GB Seagate ST9160821A PATA disk. The replaced PATA
disk (60GB Seagate ST960822A) was already a replacement of the original
40GB drive delivered with the ThinkPad, and is now used as a secondary
data disk in the UltraBay. The system is running a Debian supplied
kernel 2.6.31.5 with bleeding edge Debian/unstable packages. While there
is still a legacy WinXP partition, it is only booted once in a while to
fetch and install the latest security updates, but isn't used for almost
nothing else. So much for the scenario.

Now for the problem: I have received two warning mails by smartd
regarding a failing disk, and it pointed to an excessively high
temperature event for the newer and bigger 160GB hard disk. Caught by
surprise, i learned that there was a temporary peak of 56 Celsius
degrees at least two times, and i also noted that this disk was usually
running at around 50 Celsius degrees most of the time, albeit while
being mostly idle as revealed by iotop. The Ultrabay disk was usually at
around 40 Celsius degrees. A long running self test for the large disk
didn't show any further issue other than that temporary heat peak.

Running the machine for testing purposes in WinXP for a few hours, the
disks usually idled at 45 and 36 degrees, respectively. I downloaded and
ran the WinXP Seatools from Seagate, verifying that both disks are
perfectly fine. Trying to increase the heat in WinXP, i had to put some
load by running a full virus scan and a defrag session in parallel. But
even under these circumstances the heat level "only" reached 49 degrees.
So the expected culprit was to be found in Linux only.

As i am in holidays at the moment, i was able to invest considerable
time searching the web for possible troubleshooting hints. And i was
surprised to find out, that this issue was related to the not so recent,
but still standing click of death issues, which are mostly related to
Ubuntu Linux but apply to Linux in general. In the past i had to disable
the disk APM because of this very issue, and was not aware that, while
the excessive Load_Cycle_Count would be put on hold, this would result
in a considerable temperature increase for the affected disks.

So i experimented a bit with the hdparm -B values for both disks, and
was finally able to have them both running at around 43 and 32 degrees,
with the larger and newer drive taking the higher temperature. The
hdparm value which seems to offer the best balance between a drastically
reduced Load_Cycle_Count and the lowest possible temperature turned out
to be somewhere around 192, but i didn't experiment any further:

[root]~ # /sbin/hdparm -B /dev/hda

/dev/hda:
 APM_level      = 192
[root]~ # /sbin/hdparm -B /dev/hdc

/dev/hdc:
 APM_level      = 192

[root]~ # hddtemp /dev/hda
/dev/hda: ST9160821A: 43°C
[root]~ # hddtemp /dev/hdc
/dev/hdc: ST960822A: 32°C

I had to adapt the /etc/acpi/start.d/90-hdparm.sh script from the
acpi-support Debian package at line 35 to replace the 254 value with
192, in order to have this setting survive reboots:

[...]
	if [ "$STATE" = "BATTERY" ] ; then



	  hdparm -B 128 $dev



	else



	  hdparm -B 192 $dev



	fi



[...]

I am aware that this can not be a general recommendation, as the best
value depends on the particular drive, whether it is a PATA or SATA
drive, etc. For example, both SATA disks in my usual T61 work horse
rarely get any warmer than 36 degrees, and seem never to touch the 40
degree boundary.

I would appreciate any feedback regarding further possible temperature
decreasing measures. And yes, i already ensured that there is no dust
inside which could obstruct any airflow.

Cheers
P. *8^)