SSD

Hope you're wrong. I am going to try 2.74 and see. BIOS is dated June 2011. It should support 3T drives.

Did anyone try to install this LSI BIOS on a BR10i? I cannot find where it says it supports 3T drives. Did I miss it? BUT ... Look here on IBM's site: According to THIS VERSION HISTORY, the BR10i version 2.72 (and newer) support large drives. Enhancement - Support 2TB LUNS and greater Download here ... http://www-947.ibm.com/support/entry/portal/docdisplay?lndocid=MIGR-5077837

MyMain is reporting a cummulative speed that takes into account not only the I/O command speed, but the other commands in the telnet script. The speed on the post read is considerably less due to the checking that the data is all zeros.

I was thinking the "-t" would verify the signature. Will run an abbreviated test tonight and give you the results of the command. I suspect unRaid is having a problem with the add vs a preclear signature problem, but we'll see.

NOW he tells me . I am going to wait for the clear to finish, but I do have another 3T disk I can experiment with later.

The preclear completed successfully, and the "-t" showed it as precleared. But unfortunately it didn't work. unRAID is now clearing the disk itself. Not sure if a preclear problem or an unRAID problem. Here is what I did: - Stopped the array - Added disk to next slot (disk12) (showed with blue ball) - Clicked the checkbox to enable the start button - Clicked start button - unRAID did something for a few seconds - unRAID screen came back, disk12 still showed with blue ball, array not started - So again I clicked the checkbox to enable the start button - And again clicked the start button - unRAID began clearing the disk Here is the relevant portion of the syslog: Jun 18 09:40:46 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:40:47 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:41:25 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:41:25 Shark kernel: md: import disk12: [8,48] (sdd) Hitachi_HDS723030ALA640_MK0311YHG2TWKA size: 2930266532 (Drive related) Jun 18 09:41:25 Shark kernel: md: disk12 new disk (unRAID engine) Jun 18 09:41:49 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:41:49 Shark kernel: md: import disk12: [8,48] (sdd) Hitachi_HDS723030ALA640_MK0311YHG2TWKA size: 2930266532 (Drive related) Jun 18 09:41:49 Shark kernel: md: disk12 new disk (unRAID engine) Jun 18 09:41:50 Shark emhttp: shcmd (163): /usr/local/sbin/set_ncq sdd 1 >/dev/null (Drive related) Jun 18 09:41:50 Shark emhttp: writing GPT on disk 12 (sdd) with partition 1 offset 64 (Drive related) Jun 18 09:41:50 Shark emhttp: shcmd (165): sgdisk -o -a 64 -n 1:64:0 /dev/sdd 2>$stuff$1 |logger (Drive related) Jun 18 09:41:50 Shark emhttp: re-reading (sdd) partition table (Drive related) Jun 18 09:41:50 Shark kernel: sdd: unknown partition table (Drive related) Jun 18 09:41:52 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:41:52 Shark kernel: md: import disk12: [8,48] (sdd) Hitachi_HDS723030ALA640_MK0311YHG2TWKA size: 2930266532 (Drive related) Jun 18 09:41:52 Shark kernel: md: disk12 new disk (unRAID engine) Jun 18 09:42:23 Shark emhttp: Hitachi_HDS723030ALA640_MK0311YHG2TWKA (sdd) 2930266584 (Drive related) Jun 18 09:42:23 Shark kernel: md: import disk12: [8,48] (sdd) Hitachi_HDS723030ALA640_MK0311YHG2TWKA size: 2930266532 (Drive related) Jun 18 09:42:23 Shark kernel: md: disk12 new disk (unRAID engine) Jun 18 09:42:24 Shark emhttp: shcmd (194): /usr/local/sbin/set_ncq sdd 1 >/dev/null (Drive related) Jun 18 09:42:24 Shark emhttp: writing GPT on disk 12 (sdd) with partition 1 offset 64 (Drive related) Jun 18 09:42:24 Shark emhttp: shcmd (196): sgdisk -o -a 64 -n 1:64:0 /dev/sdd 2>$stuff$1 |logger (Drive related) Jun 18 09:42:26 Shark kernel: sdd: sdd1 (Drive related) Jun 18 09:42:26 Shark emhttp: re-reading (sdd) partition table (Drive related) Jun 18 09:42:26 Shark kernel: sdd: sdd1 (Drive related) Jun 18 09:42:27 Shark emhttp: clearing disk12... (Other emhttp)

I ran preclear in a new window with the -t option (didn't think it would hurt the running post read). Looks like everything is okay. Guess won't know for sure until I try to add the disk to the array. ################################################################## 1.12 Device Model: Hitachi HDS723030ALA640 Serial Number: MK0311YHG2TWKA Firmware Version: MKAOA3B0 User Capacity: 3,000,592,982,016 bytes Disk /dev/sdd: 3000.6 GB, 3000592982016 bytes 255 heads, 63 sectors/track, 364801 cylinders, total 5860533168 sectors Units = sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disk identifier: 0x00000000 Device Boot Start End Blocks Id System /dev/sdd1 1 4294967295 2147483647+ 0 Empty Partition 1 does not end on cylinder boundary. ######################################################################## ========================================================================1.12 == == DISK /dev/sdd IS PRECLEARED with a GPT Protective MBR == ============================================================================

I realized since I was in the post read step, that the disk was already precleared, so decided to see if "gdisk" recognized it as a valid GPT disk. See below. Is this what I should be seeing? I am hoping that running this command didn't mess up the preclear signature. I did run the command a second time and got the exact same results, so am hopeful that the "converting MBR to GPT format" did not write anything back to the disk. Does the "-t" option of preclear work for large disks in this beta version? >gdisk -l /dev/sdd GPT fdisk (gdisk) version 0.6.14 Partition table scan: MBR: MBR only BSD: not present APM: not present GPT: not present *************************************************************** Found invalid GPT and valid MBR; converting MBR to GPT format. *************************************************************** Disk /dev/sdd: 5860533168 sectors, 2.7 TiB Logical sector size: 512 bytes Disk identifier (GUID): A624A0B1-9F34-4185-8DEE-E6985C5EED53 Partition table holds up to 128 entries First usable sector is 34, last usable sector is 5860533134 Partitions will be aligned on 2048-sector boundaries Total free space is 5860533101 sectors (2.7 TiB) Number Start (sector) End (sector) Size Code Name

Running at 143 MB/sec now @ 7% complete. Not sure I'll be home when it gets to the inner cyl, but will have to see. BTW, this is a 7200 RPM Hitachi drive. Write phase finished in about 7 1/2 hours. I wasn't watching at the end, so don't know speed on inner tracks. Working on post read now.

Running at 143 MB/sec now @ 7% complete. Not sure I'll be home when it gets to the inner cyl, but will have to see. BTW, this is a 7200 RPM Hitachi drive.

Yes, I have just finished one... I'll attach it to this post temporarily. (Until it gets tested by a few brave users) It has NEVER been tested on an actual 3TB drive... It should work. The GPT signature it creates is identical to the one Tom at lime-technology described to me. You'll have to give it a try and let me know how it goes. It can not convert GPT partitions using the "-C" option, as they are handled entirely differently and never start on partition 63 or 64. Also, if your disk is over 2.2TB in size, it will set the preclear signature as it needs, completely ignoring any "-a" or "-A" option you might try to give it. Joe L. Running now, with the "-W" option (disk was precleared previously and has been running in 2.2T mode for a while). Zeroing at 140 MB/sec The initial screen is still saying stuff about partition alignment / offsets, etc. It should probably say it is a >2TB disk and being prepared for a GPT structure. My disk already had an MBR from the 2.2T mode (all I did was remove the HPA, reboot, and start this preclear). So this should be a pretty good test of preclear's ability to preclear a 3T drive.

I actually prefer that the Supermicro does not have power lights for each slot. When disks are being acessed it is very obvious. Figuring out which slot is occupied is not a biggee for me.

Nice thread on where to shop for UPSes and replacement batteries: http://lime-technology.com/forum/index.php?topic=12548.0

Can you elaborate a bit on what make you feel this way? Why certainly good sir. ICY Dock issues where mostly around what I perceived to be cheap drive cages. To me, this did not justify the cost. Right now it is $25 more then the next two cheapest, the Norco and SM units. That along with what lookd like not the greatest airflow in the rear made me less than thrilled. Once I plugged in my Corsair Molex cables and then tried to plug in SATA cables, quickly realized that wouldn't work. Many of the modular Molex cables come with those tabs on them (here) and those won't work. Icy Dock layout places some of the SATA plugs right below the power plug. No other cage had this design. So have to go find a splitter... Don't get me started about the fan either. iStarUSA - cheapest of the bunch, does have that going for it. But on the model I have at least, zero airflow openings in the front. Be really concerned with heat. I cannot test right now as I have already had to RMA the unit due to tray5 not working. I was turned off on the design of the locking mechanism for the drive cages. It seemed like the slightest pressure on them would pop it open, forcing me to use the itty bitty locking mechanism. Which is probably there because they know they pop open so easily. Unlike the Norco and SM, another custom fan connector on this one as well. After working with all four, I can say these two are in the bottom of the pack. But, this is my opinion. I really liked working with the Norco and SM models. Now, at $95 you cannot beat the price for the iStarUSA unit and both it and the Icy Dock will more than meet most people needs. But, personal preferences and all, I would go for the other two. Shawn Very helpful! Will be interesting to see your temperature benchmarks and see which ones do the best job keeping drives cool. Thanks Shawn!

Monoprice.com Monoprice is a great place for cables (SATA, network, audio, video, and more), splitters, adapters, and a few PCI / PCIe controller cards. They do not carry the SAS breakout cables, but there was a recent post that they might be adding them. This vendor has an excellent reputation for quality products at low cost.

Can you elaborate a bit on what make you feel this way?

After using both, for my purposes I like the SuperMicro better. The larger fan and the aluminum trays trays are all positives for me. (I have about 50% 7200 RPM drives.) I do not think that the Norco tray bottom would act as a heatsink. A heatsink requires direct physical contact to work very well, a good heat conductor (copper or aluminum most commonly), and a shape to help disapate the heat. I think the bottom would have the opposite affect, it would trap hot air between the bottom and the drives and interfere with ventilation. I think I would dremel them out if I had one. I also wanted to comment on the plastic handle. I agree it feels a little flimsy, but I have never heard of anyone breaking one. It does make you treat is a little gently, which might be a good thing, so that you are less apt to try and force the thing closed when it isn't in quite right.

I believe the Norco trays are steel.

This is awesome! Thank you so much for taking the time to do this review. A few comments, based on some experience with these cages: 1 – The SuperMicro is the only cage with a 92mm fan. Other all have 80mm. This has an impact on the cooling potential of the cage. So long as the drive temps are controlled, the fan size doesn’t matter. But different seasons might require more or less cooling. 2 – The hottest part of the drive is the bottom. Having the “cage” more open on the bottom allows more airflow. Both the SuperMicro and the Icy Dock are completely open on the bottom. The iStar is more ventilated than the Norco. (I could see a dremeling opportunity with the Norco cage bottoms). 3 – Although the SuperMicro is the deepest of the bunch and has caused problems for some cases, the Norco unit may create challenges for some. The SuperMicro clipon fan is not as tall or wide as the main unit, this leaves room for cable connections and for motherboard overhang. When considering your case and cage preference, keep the depth as well as height and width towards the back of the unit, in mind. 4 – I have had both the Norco and SuperMicro. Both are well made. The nice thing about the Norco is that it does not have these removable plastic inserts (the size of a hard disk) that must be reinserted (with the annoying screws) should a disk be removed. (Presumably the plastic insert maintains similar airflow characteristics with the plastic insert in place). But the SuperMicro has the completely open top and bottom. Looks like the IcyDock have just a bar that must be removed / replaced with a disk is inserted / removed. Similar concept. Although the SM plastic trays are a pain, they do make good spare part bins . 5 – Interesting that the Norco is the only unit that has individual air channels for each drive. This means less room for airflow between the drives, but also, in combination with the air shutoff feature on each cage, allows you to block airflow through certain channels. The jury is still out, IMO, whether the individual channels are a positive or negative. I am really interested in the following test. Load all 5 slots with disks (5 x 5400RPM). Disconnect all other cages (so they don’t impact the test). Close up the case, and run preclears (with –n option to immediately begin writing which should generate the most heat) on them for a certain period of time (say 1 hour). Check all of the drive temps. Do it in each of the cages (same disks in same order), running just drives in that cage. Then repeat in the next cage. Repeat with 7200RPM disks. You could run this with a dd command from disk to /dev/null (if you do this use a larger buffer). I don’t think drives get quite as hot when reading, but results would likely be similar if actual writing to the disk is not desired. Report ambient when you report your findings, in addition to your case cooling. A slightly warm ambient might be most useful to give a sense of how hot the drives would get under summertime conditions (ambient about 78F).

Buy.com Buy.com is a great place to shop for the SuperMicro CSE-M35T-1 5-in-3 "hotswap" disk cages. At ~$96 with free shipping, I have not found better. There is an easy to find $5 off first order over $100. Also, budget (free) shipping is amazing. I placed an order on Friday and it arrived by 10AM the next day (Saturday morning). Last time I ordered from them the order also arrived next day (but it wasn't a weekend), and I thought it was a fluke. Shipping for this item was from PA, which is about a 5 hour drive from where I live. After 4 perfect transactions from buy.com, I would recommend them.

We all know that Newegg and Amazon are good sources for unRAID parts and drives. But there are other merchants that are also well-priced and worth considering. And there are certain places that are the best for buying specific things. The purpose of this thread is to allow users to provide specific recommendations on where to (or not to) shop for what.

12 at a time might be a record! Are the speeds > 100 MB/sec? Will be interested in total wallclock time to complete. How much RAM is in your machine?

Any theories of why this would happen? I'm wondering if these were aligned in time. Might be nice if preclear logged the date/time of a preclear post verification read error. And even do some retries to see if it is really a data issue, or a memory or other system issue.

I got some strange results running preclear (v 1.6, unRAID version 5.0b4). I was running three preclears at once in screen - 2 3T and 1 2T drive. Each of them got a single error on the verification step. Here are the results of each of the postread_errors files. skip=236000 count=200 bs=8225280 returned instead of 00000 (3T) skip=260800 count=200 bs=8225280 returned instead of 00000 (3T) skip=184400 count=200 bs=8225280 returned instead of 00000 (2T) This is the first time I have run a preclear on this (new) server. It is running a C2SEE-O motherboard (2 3T drives on this controller), and a BR10i controller (2T drive is on this controller). After a reboot I tried to double check that the values of these blocks was truly zero. Here are the results: root@Shark:/boot/unmenu# dd if=/dev/sda skip=236000 count=200 bs=8225280 conv=noerror 2>/dev/null|sum| awk '{print $1}' 00000 root@Shark:/boot/unmenu# dd if=/dev/sdb skip=260800 count=200 bs=8225280 conv=noerror 2>/dev/null|sum| awk '{print $1}' 00000 root@Shark:/boot/unmenu# dd if=/dev/sdd skip=184400 count=200 bs=8225280 conv=noerror 2>/dev/null|sum| awk '{print $1}' 00000 Whatever the problem, it is not memory. I ran the memory (4G) through a 2 night memtst - no errors. Have you ever seen this before? Any ideas what might have caused it?

How many read-only parity checks did you run and get the same results? Based on my read of the thread, I think you got 12 the first time, then again and got 11, and then a third time and got 11. Is that right?