topic Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating in Archive

320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Mon, 25 Apr 2011 16:25:38 GMT

Just installed 5 of these in a RAID configuration on a test server. Working perfectly, except:

The server is reporting that the drives are overheating (they're not). It appears that possibly I can turn OFF DIPM on these drives and the SMART info will be reported correctly.

How do I do that?

FYI, running windows 2008 r2

Thanks,

Rob

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Tue, 03 May 2011 05:22:20 GMT

Let me check if Intel get the email.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Thu, 05 May 2011 21:08:14 GMT

Any response from them?

Thanks,

Rob

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 06 May 2011 08:05:15 GMT

we are facing the same issue.

greatly appreciate a rapid solution.

thanks!

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 06 May 2011 15:50:54 GMT

All,

I'm working to get you an update. Haven't forgotten about this

-Scott, Intel Corporation

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 06 May 2011 15:53:59 GMT

Thanks!

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 06 May 2011 18:56:56 GMT

Funny thing is.... I have a procurement order for DL380/G7 myself at work!

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Wed, 11 May 2011 07:36:09 GMT

Same problem here. Has lodge a case with Intel support center. Still waiting for feedback.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Thu, 12 May 2011 16:04:21 GMT

How do you lodge a case with Intel Support?

Thanks,

Rob

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Mon, 16 May 2011 19:05:21 GMT

Same problem here!!

Tested in a HP dl380 g7 performance model whit

1. INTEL 510 SERIES 2.5" 120GB SSD SATA/600 MLC 34NM

2. INTEL 320 SERIES 2.5" 120GB SSD SATA/300 MLC 25NM

3. INTEL 320 SERIES 2.5" 160GB SSD SATA/300 MLC 25NM

The all say overheated, and the server makes a terminal shutdown

The logs say 250 degrees Celsius in disk bay (That's quiet hoot)

Seems that it's the G3 model that doesn't work in HP dl 380 g7 servers

G2 models works perfectly in all my HP dl 380 g7 server

But the G2 are really difficult to find in the market today, only G3 in the shops now a days

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Wed, 18 May 2011 21:16:32 GMT

I also have the same results.

HP DL380 G7, Intel SSD 320 600GB.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Thu, 19 May 2011 11:31:48 GMT

For those technical: I can speculate exactly what the problem is, and in my opinion HP needs to deal with the issue, not Intel.

Chances are, whatever monitoring software HP provides -- or, if it's the system BIOS complaining, then HP's BIOS -- is making a horrible assumption that the disk attached is a classic MHDD. Most MHDDs -- but not all -- provide SMART attribute capability that includes SMART attribute 194 (0xc2), which is commonly used for temperature. This is the only way I know of (and I'm http://bsdhwmon.parodius.com/ quite familiar with hardware monitoring and http://jdc.parodius.com/freebsd/atacontrol/ quite familiar with SMART) to obtain hard disk temperature. Some drive vendors also track minimum, maximum, and average temperatures seen (possibly they tie this into hard disk SCT capability; unknown).

Anyway, Intel SSDs don't provide any data for SMART attribute 194; specifically, the attribute doesn't exist on Intel SSDs. A non-existent attribute is 100% normal (it's not the same thing as "it has the attribute but related data is zero").

So, the HP software (or BIOS) may therefore be buggy/broken; it seems they may be making the assumption that attribute 194 exists on all drives installed in their ProLiant systems, but that simply isn't true. If this is true, then their code is simply wrong and needs to be fixed. Again: if this is an error that either happens during or immediately after system POST, then HP needs to fix their BIOS.

I think it's a safe assumption HP *is not* doing something like using a Winbond or LM-series chip to monitor hard drive bay enclosure temperature (nothing would report 250C if that were the case). My above theory is much more plausible.

If someone technical (read: not someone saying "derk derk my system says the temperature is high!! fix it!!" -- I'm talking about an engineer) can provide full details of **what** is reporting an excessive drive temperature, maybe some light can be shed on **how** the software is obtaining that information. I sure hope HP provides full technical documentation of whatever their software is monitoring. In the case it's their BIOS, then they should be ashamed.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Thu, 19 May 2011 15:57:14 GMT

While this may be true, why did prior generations of Intel SSDs not have this problem reported, and why do other SSD manufacturers not have this problem?

Is it not possible, ever, for an Intel SSD to overheat?

To maintain consistency and to get a fix sooner, I would hope that Intel would choose to report this data point, and either report it correctly, or at least report it with a "neutral" value.

Otherwise, HP would have to maintain a table of SSDs, and which one reports "what".

Rob

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Thu, 19 May 2011 22:47:54 GMT

To All,

The Intel 320 and 510 Series Drives do not contain a Temperature Sensor. Intel 320 and 510 series drives are responding within the proper limits of the SATA specification for drives without temperature sensing capabilities. We are working closely to help HP better understand the issue. Thanks,

Scott, Intel Corporation

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 20 May 2011 07:51:18 GMT

I state the below under the assumption that HP is requiring (and incorrectly assuming that such exists) SMART attribute 194 to determine temperature of the drive:

As a senior UNIX system administrator (read: I work in an enterprise environment) who's also well-versed in ATA specs, I strongly disagree with the opinion that Intel should violate specification and return "fake SMART attributes" just to keep some particular model of chassis/mainboard or piece of OS software happy. There is *absolutely nothing* that requires a manufacturer to implement SMART -- and if they do, there is *absolutely nothing* that mandates a pre-requisite list of SMART attributes that must be implemented. Period. This isn't hearsay, it's fact. If anyone would like to read the ATA-7 (production) and ATA-8 ACS-2 (working draft) specifications, head on over to t13.org and read 'em. SMART is 100% optional and what attributes a drive vendor implements is completely and entirely up to them. If my theory is correct, then HP *requiring* a drive to provide a thermistor tied into SMART attribute 194 is absolutely 100% the fault of HP and not the fault of Intel. I can absolutely see a company who has a long-term history of mandating use of their "own hard disks" (Sun, HP, Compaq, IBM, etc. -- all of whom use rebranded Seagate, Hitachi, or Fujitsu drives) making this sort of assumption. Furthermore, most SSDs (truly!) do not provide a temperature thermistor -- it's not limited to just Intel.

I'm very glad Intel is working with HP directly to solve this problem, and I look forward to seeing what the root cause is. In my opinion, customers should be happy that Intel is assisting HP to rectify this issue. My advice to ProLiant customers would be to put serious pressure on your sales reps to get something done about this. You have support contracts for this exact reason.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 20 May 2011 13:25:34 GMT

Thanks for the update. It does sound like it is in HP's court.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Fri, 20 May 2011 16:12:00 GMT

I have 4 Intel 320 series 120GB drives with the same problem in a HP DL385G7. I also have several DL380G6's using 12 of the 160GB G2's.

The 320 series is supposed to be a direct replacement of the G2's, and indeed share the same firmware updates.

On the machines running the 160's, no issues. The second I pop the 320 series in, I get the fans spinning up, and the iLO begins to shut down the server because it thinks the drives are about to go nuclear.

Here's the problem:

The server in question has 8 bays. The first 4 have WD 600GB velociraptors in there. This is how they report temperature according to SMART.

Physical Port 1I

Physical Box 1 (0x01) Physical Bay 4 (0x04) RPM 10000 RPM (0x00002710) Device Type Sata (0x01) SATA Version 0x00 Big Total Block Count 0x0000000045dd2fb0 RIS Starting LBA 0x0000000045dc37b0 RIS Size 160 KB (0x00000140)......

Vendor Unique Inquiry Bytes All Zeroes (20 x [0x00])

Current Temperature 0x1c Temperature Threshold 0x37 Maximum Temperature 0x2d

This is the same log file for the 320 series:

Physical Port 2I

Physical Box 1 (0x01) Physical Bay 5 (0x05) RPM 0 RPM (0x00000001) Device Type Sata Ssd (0x01) SATA Version 0x00 Big Total Block Count 0x000000000df94bb0 RIS Starting LBA 0x000000000df853b0 RIS Size 160 KB (0x00000140)......

Vendor Unique Inquiry Bytes All Zeroes (20 x [0x00])

Current Temperature 0xff Temperature Threshold 0x37 Maximum Temperature 0x00

So, Intel's 320 drive is reporting current temp as 0xff (max), and a max temp of 0x00 (min). If the previous poster was correct, and the drive simply reported 0x00 as it should for unsupported features, then there would be no problem. This is NOT what it is doing however. I've been waiting with an open ticket for a MONTH for a resolution on this. I don't care if it reports a fake number, I don't care if it reports 0x00. I just want to use the friggin $1000 worth of SSD's for my VM's like I'd planned on and budgeted for.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Sat, 21 May 2011 03:37:40 GMT

Aaron, your conclusion is incorrect in numerous regards, and it's because you don't understand how SMART attributes are stored nor are familiar with the protocol. It's understandable (most people aren't), and your frustration is justified.

First and foremost, as I've stated twice already, Intel SSDs do not have any support for SMART attribute 194. There is no thermistor. The SMART attribute isn't even listed/provided by the drive when it responds to ATA feature command 0xd0 (obtain SMART attributes and their data). So, simply put, there is no 0xff value for the iLO (or whatever; BIOS, firmware, etc. -- doesn't matter to me what it is) to obtain. Secondly, each SMART attribute consists of 6 bytes of data, not 1 byte.

Here is the SMART data structure that's returned from ATA feature command 0xd0. This is taken from my own SMART code for FreeBSD's atacontrol:

148 /*

149 * Obtain SMART attributes and their values150 *151 * Feature 0xd0 result (see ATA8-ACS, section 7.53.6.2, table 49):152 *153 * The 512-byte result of SMART READ DATA is documented per ATA8-ACS154 * specification, section 7.53.6.2, table 49. However, you'll find155 * bytes 0-361 marked "Vendor specific"; these are (mostly) the156 * actual SMART attributes themselves. Example:157 *158 * Offset Size (B) Description159 * -------- --------- -----------160 * 0 2 SMART attribute revision (16-bit, big endian)161 * 2 12 SMART attribute data entry # 0162 * 14 12 SMART attribute data entry # 1163 * .....164 * 348 12 SMART attribute data entry # 29165 * ..... 166 * -------- --------- -----------167 *168 * The SMART attribute data format is completely undocumented. It169 * consists of 12 bytes per attribute in the following format:170 *171 * Offset Size (B) Description172 * -------- --------- -----------173 * 0 1 Attribute ID number174 * 1 2 Attribute flags175 * 3 1 Attribute CURRENT value (adjusted)176 * 4 1 Attribute WORST value (adjusted)177 * 5 6 Attribute data178 * 11 1 179 * -------- --------- -----------180 */

Simply put, the iLO device *is not* getting back 0xff from the SSD itself. There is probably a broken piece of code in their iLO firmware or iLO BIOS which makes the assumption SMART attribute 194 exists, tries to refer to it in some in-memory buffer, and gets back whatever the contents of that buffer are (which haven't been populated/filled because there's nothing there). Lots of buffers, especially on embedded hardware (which the iLO is considered), are initialised with value 0xff rather than 0x00. There's a lot of reasons for this initialisation value, and the value chose IS NOT the bug. The bug is almost certainly that HP *assumes* all drives installed in their chassis support SMART attribute 194, and that is absolutely wrong/false/broken.

Your Western Digital WD6000BLHX or WD6000HLHX drives do have thermistors, and WD populates SMART attribute 194 with appropriate thermistor data.

So again: please send all of your complaints/concerns to your support reps at HP, because this situation is exactly why you pay for support contracts. The more pressure you put on HP the better. There's absolutely nothing that requires a hard disk vendor (mechanical or SSD; doesn't matter which) to support SMART attribute 194, nor do they have to install a thermistor in their drives at all. HP therefore is making a very bad assumption, and the only persons suffering from it are their customers.

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Mon, 23 May 2011 07:23:14 GMT

Will it solve the problem if adding a Intel RAID controller?

Re: 320 / 600 GB in Proliant DL380/G7 - shows as overheating

idata — Tue, 24 May 2011 12:31:03 GMT

This is aN HP error.

We use Non intel SSD and get the same issue.