Replacing disk drive firmware

I have performed EMC Clariion Seagate OEM disk firmware upgrades with Seagate generic disk firmware ( *.lod associated with disk type ). The disks firmware revision changes, but the disks type label remains the same ( CLAR18G Solaris or ST318203FC Seatools ). This would lead me to believe that what ever firmware is factory loaded onto the disk remains the same and any patches applied to the disk are just updates.<b>*</b> Removing the firmware and then reapply the new firmware is not possible with current tools available ( Seatools, fwdl and format ). Also you will render the disk useless if the upgrade procedure goes wrong ie. if you install incorrect firmware. The disks firmware plays a vital role in the disks operations, without it the the disk will not function ( removing the firmware will prevent access to the disk and therefore you will be unable install the new firmware ). It would be like removing open boot prom software or PC BIOS, the only way to get the system back is to install a new prom chip. How are you trying to upgrade the firmware ( what tools are you using ) and why?

<b>*</b> After a little research on this topic, I have updated my post.

mlennon at 2007-7-5 17:50:02 >

I understand why someone may wish to perform firmware upgrades, but that's what it is, an upgrade. Most of the tools available for disk firmware modifications merely apply patch software to an existing image, the tool analyzes firmware *.lod description against the installed drive, if there is no match it will fail. Modifications to nStor systems using nStor tools should be discussed with nStor support. This forum is for supporting Sun hardware related issues, so on the topic of disk firmware on Sun disks this pertains to firmware upgrades, <i>./download</i> tool for photon for example and not modifying Sun components to be used in other vendors equipment.

mlennon at 2007-7-5 17:50:02 >

The issue with the firmware has never come up and this is why I was interested in your post, I have been experimenting with various firmware revisions and different OEM drives, but mainly EMC Seagate drives and Seagate standard drives. If you have a SAN environment and large numbers of spare Sun drives perhaps you should create a JBOD storage area using an A5200 and manage the disks using an older Sun server, I have recycled drives this way and it provides reasonably large network storage for PCs.

mlennon at 2007-7-5 17:50:02 >

It was me that made that statement, I didn't say " because they didn't have a Sun firmware on them... " I said that these drives were not developed for Sun firmware controllers and therefore are not supported.

For anyone that wants to read that discussion here is the URL:

<a href="http&#58;&#47;&#47;supportforum.sun.com/hardware/index.php?t= msg&amp;goto=19533&amp;rid=3131&amp;SQ=a65a2486061248900367e146cde90 749" target="_blank">http://supportforum.sun.com/hardware/index.php?t=msg&goto=19533&rid=3131&SQ=a65a2486061248900367e146cde90749</a>

Disk spindle speed only plays a part in disk performance, so just slapping in any high speed disk into a system that it has not been optimized for isn't going to magically yield a massive performance increase you seem to expect. Your statement " performing like dogs " really proves my point. Your Compaq disks are designed to run with different controllers, Compaq developed controllers with Compaq firmware installed ( probably developed for Windows environment ). There are many many factors that contribute to I/O performance. One of the biggest issues facing implementations of large heterogeneous arbitrated loops is interoperability between different vendors products, this also applies to mixing different vendor drives or firmware revisions within a JBOD. Each device within an FC JBOD array or taking part of a FC_AL, devices upstream transmitting to ports on devices downstream, are dependent on the device firmware to control each devices behavior within the loop. Fibre channel protocol is made up of 5 layers, layer 2 ( FC-2 ) is where firmware controls flow of data. Each manufacturer optimizes it's firmware to enhance interoperability and performance at this layer and this is the reason why companies like Compaq, EMC, IBM and Sun develop versions of firmware for these OEM devices ( optimizing devices to work together ie. RAID controllers, HBAs and disks ). Sun machines that are to be attached to an EMC or IBM storage systems will need revised firmware for the HBA to operate efficiently with the storage systems controller. So in other words any mismatch in software at FC-2 layer will greatly degrade loop performance. Now back to the topic that this thread was started on, firmware is not completely erased from the disk with any of the tools I have mentioned ( Sea tools, fwdl or Solaris download ). These utilities merely update the installed firmware. These utilities depend on information embedded into the disk firmware and any mismatch will cause the application to exit. Here are examples of this information gathered from different versions of Seagate drives ( Generic and Sun ) using Sea tools:

<b>Genereic Seagate</b>

Vendor = SEAGATE

Product = ST318203FC

Version = NA12

Serial Number = LR547324

Copyright = Copyright (c) 2003 Seagate All rights reserved

SCSI Firmware = 0221DA12

Servo RAM Release = B34AB34A

Servo ROM Release = 00010001

Servo RAM Date = 0000

Servo ROM Date = 0000

Blocksize = 512, Highblock = 35566479, Capacity = 17783 MB

-this is a Seagate drive

-this drive does not support DST

-Mode Page Settings [current value (default)]:

-WCE bit = 0 (0)

-RCD bit = 0 (0)

-AWRE bit = 0 (0)

-ARRE bit = 0 (0)

-DExcpt bit = 0 (0)

<b>-Number of cache segments = 16 (16)</b>

-DLM bit = 0 (0)

-DTOLI bit = 0 (0)

<b>SUN Seagate</b>

Vendor = SEAGATE

Product = ST318203FSUN18G

Version = 034A

Serial Number = 0030J586

Copyright = Copyright (c) 1999 Seagate All rights reserved

SCSI Firmware = 07120403

Servo RAM Release = B34AB34A

Servo ROM Release = 00010001

Servo RAM Date = 0000

Servo ROM Date = 0000

Blocksize = 512, Highblock = 35378532, Capacity = 17689 MB

-this is a Seagate drive

-this drive does not support DST

-Mode Page Settings [current value (default)]:

-WCE bit = 0 (0)

-RCD bit = 0 (0)

-AWRE bit = 1 (1)

-ARRE bit = 1 (1)

-DExcpt bit = 0 (0)

<b>-Number of cache segments = 2 (2)</b>

-DLM bit = 0 (0)

-DTOLI bit = 0 (0)

So if Sea tools application is instructed to update the firmware on either of the above disks it attempts to match the header information of the disk to the header information of the firmware upgrade image, if there is no match the application exits. Solaris download is even more strict in it's operation, if it does not detect an attached Sun storage system it exits immediately.

The Sun version this drive ( ST318203FC ) is also supported on the Sun Blade 1000, brought to the discussion by my Californian friend who, along with his Sun experienced people, contradicts my explanation why non Sun disks don't perform well in Sun machines. Above we see in bold that the Generic disk drive that has been optimized for use in a Windows environment has 16 cache segments and the Sun version has 2 cache segments. Generally less disk cache segments are desired for high performance UNIX servers, workstations and storage systems ( disks optimized for other high performance environments always use less cache segments as well eg. EMC Clariion, SGI Irix and Apple OS use 3 segments per cache ) because high performance systems need large amounts of data at sustained transfer rates as quickly as possible. So in order to run these generic disks in a high performance UNIX workstation, the disk must have revised firmware. Seagate do not offer firmware to the public and it is generally distributed through the OEM who optimizes it for use with it's storage products. On an almost weekly basis, I have customers ask me " why should I buy Sun storage hardware when I can pick up the generic disks from Dustin ( or some other PC reseller ) for a fraction of the cost? ", years of R & D have gone into refining and optimizing disk firmware by many of the high performance storage manufacturers, this development is what you are paying for - data integrity and availablility. If you compare any of the enterprise class storage systems currently available on the market, all carry huge price tags on EOL FRUs like disks. Sun EOL systems components carry big price tags because that option is no longer in production and therefore is only available in reserve. For customers that need to extend the life of hardware, Sun offers the remanufactured program, where EOL options are available for less than half the list price carrying a full as new warranty. Sometimes these parts have a long lead time, so in that case there are plenty of used resellers offering used Sun disk options with a limited warranty for under $500. So there is no reason to go and buy unsupported PC or other OEM parts for your Sun system. Alternatively there are some relatively cheap low end SATA storage systems available with RAID controllers that may be a better option for a UNIX shop that need space but don't have the budget for higher performance systems. Sun resellers have some cracking deals available on 3511SATA array at the minute.

mlennon at 2007-7-5 17:50:02 >

Just remeber that if your disk is segmented in the way I have described and if you are reading large amounts of data concurrently your application will be reading from the disk all the time, this will put a heavy load on the system. There are expert options in format, but I would not recommend playing around with it unless you know what you are doing. If I remember correctly it only allows you to enable or disable the cache, disable the cache may make things even worse. You could always ask HP for firmware upgrade, but I would imagine that it is only available to contract customers that have bought a complete storage system. D1000 is a good system if you need lots of disk space but I would think that is has poor performance, I think the max transfer rate is 20 - 40 MB/s ( probably much less ). If I were in your shoes I would look at a T3 instead, might cost more but it will have much better performance, here is an interesting thread on the system:

<a href="http&#58;&#47;&#47;supportforum.sun.com/hardware/index.php?t= msg&amp;th=5105&amp;start=0&amp;rid=3131&amp;SQ=8aa2e11b85e15e33 d67c84e78df94ab2" target="_blank">http://supportforum.sun.com/hardware/index.php?t=msg&th=5105&start=0&rid=3131&SQ=8aa2e11b85e15e33d67c84e 78df94ab2</a>

On the difference of opinion with the other Sun people who said that my statement about the firmware was questionable - If only it was as simple as just using any Fibre Channel components from any system! I would take, with a pinch of salt, any suggestion from anyone that recommend using components other than items recommended by Sun. If you look at the lengths that storage manufacturers ( EMC and IBM for example ), go to to develop firmware for controllers that are to be used in different environments ( AIX, HP-UX, Solaris, Linux, Apple OS and Windows ). For each environment the firmware has been carefully optimized, this isn't some kind of money making scam - relabeling products and selling them on for twice the price, there's quite a bit of development going on in the background!! I have almost given up completely on trying to explain this to some of my customers!

BTW I am from Ireland, I just live in Sweden!

mlennon at 2007-7-5 17:50:02 >

These drives that I used here for demonstration purposes were decomissioned from a bunch of Blade 1000s systems. The drives were upgraded to 36GB items and are currently not in use. If I was to recycle these drives into a photon array or a system, the patch matrix for the system would be applied prior to the system going into production or being sold to a customer, otherwise the disks would sit here on a shelf doing nothing ( gathering dust if you will! ).

mlennon at 2007-7-5 17:50:02 >

This post does answer Johans question, but there are a few things that I would be concerned about here. But a few questions first: " make up some story about you get a reload-firmware error.. " are suggesting that we contact seagate with some b*&$sh1t story? This is not something that I would suggest to my customers, dishonesty in any way doesn't go down well in business. As I mentioned in my previous posts, if the firmware upgrade goes wrong you render the disks useless, this is not a risk I would be willing to take. I deal with small amounts of used disks and I actually sell the generic seagate drives cheaper than the Sun versions. Furthermore, the discussion that unixmuseum and I have been having, is a kind of proof that using disks that have been designed for use with a PC environment ( generic releases ) for heavy I/O applications actually degrades performance.

<b>NOTE</b> For the record, I made a reference in this post stating that I sell used disks. In no way was this statement an attempt to offer used hardware through this forum. I made to comment as example of the current market rates for Seagate euipment both generic and Sun OEM. So in other words the Generic disks can be bought cheaper than the Sun versions from any vendor.

m-lennon at 2007-7-21 14:28:21 >

I would have no objections if you wanted to remove any of the entries I have made ( sorry about the remark in my last post I will edit it, though I was trying to express my thoughts on the suggestion ). I feel that the thread has moved beyond the forum description of Sun hardware support and into discussing OEM manufacturer released tools that aren't even available for Solaris. In my previous posting I used Seatools to gather information from the drives for demonstration purposes only, format can get this information from the disk also, but it is a little cryptic to understand, so I thought that Seatools output was simple.

m-lennon at 2007-7-21 14:28:21 >