PeterB

Hmmm ... it appears that I spoke too soon. After 17 hours, when the second drive was on its Post-Read, I'm starting to get: Oct 30 10:09:21 Tower kernel: sd 0:0:1:0: attempting task abort! scmd(f5a8d540) Oct 30 10:09:21 Tower kernel: sd 0:0:1:0: [sdh] CDB: cdb[0]=0x85: 85 06 20 00 00 00 00 00 00 00 00 00 00 40 e5 00 (Drive related) Oct 30 10:09:21 Tower kernel: scsi target0:0:1: handle(0x000a), sas_address(0x4433221101000000), phy(1) (Drive related) Oct 30 10:09:21 Tower kernel: scsi target0:0:1: enclosure_logical_id(0x50030480041f4c00), slot(2) (Drive related) Oct 30 10:09:21 Tower kernel: sd 0:0:1:0: task abort: SUCCESS scmd(f5a8d540) Previously, I would see these messages within seconds of starting the Pre-clear. However, this is on a suspect drive, with a transfer rate falling to less than 7MB/s. I've already discussed this drive in the past. SMART shows nothing untoward - the only symptoms I see are this dreadfully slow transfer rate at around 70-80% through the Pre-clear reads and, I guess, these strange error messages in the log. To RMA the drive (a WD10EADS), I would have to send it internationally from Philippines to Singapore.

... and I thought that I was the only person experiencing those errors! My Supermicro AOC-USAS2-8Li started exhibiting the same problems when I added the iStarUSA cages into my build - parity checks would fail (although rebuilds were okay) with this error, and preclears would fail within seconds of being started. I have just updated to the Phase 11 firmware, and first a single, then two simultaneous, preclears appear to be running without a problem. I never had any segfault problems with 12a, but reverted back to beta 11 because of nfs errors in 12.

I'm using that Seasonic X650, and love it! The reason that it's on offer is that it's been replaced by the X660. However, the 660 comes with only five 'molex' peripheral connectors (eight on the 650). This may, or may not, be a problem for you, but my plan to add 5x3 cages to my machine dictated that more molex would be required. This was enough to swing me towards the X650!

The iStarUSA BPN-350V2 has a blue led for every drive which is installed/powered. Together with the blue 'power on' led on my Thermaltake case, it all looks very cool!

Does this really remove the -d ATA option which unMENU adds to each smartctl command? It is this which causes problems with many add-on controllers. I'm still not clear why it was felt to be a good idea to force smartctl to attempt to treat every drive as an ATA device. Although this appears not to upset some SATA controllers, others refuse to 'play' when they're treated as a legacy ATA device. I have hand-edited every file in unMENU, where necessary, to remove the '-d ATA'.

Just a few simple patches to the script files will fix this.

Really? You surprise me! ... but I'm convinced that trayless is the way to go! As for temperatures - I've had a rebuild running overnight - now, at 8 am, ambient is 27C. Highest drive temperature is 36C and no noisy cage fans - I'm pleased with that.

Which is why I think that it was very public-spirited of you to conduct the comparisons. Not only do you have to put up with using trays, but you have four incompatible tray types to contend with!

It's your build... & not mine Thanks for the compliment! I do have to say that I'm very pleased with it ... with, perhaps, a little pride creeping in!

PATA drive? Yep, if you look carefully at the connector, you can see that it's PATA, not SATA! I'd just fitted the cage into the case and had a pile of old PATA drives on the table beside where I was working. The SATA drives were across the other side of the room, and I'm lazy!

Thanks! The drive is inserted the wrong way! Nope, the drive is the correct way round - electronics to the left, connector to the rear.

Sure, but the Thermaltake case does make it relatively relatively easy to seal up. I have two 120mm fans and one 170mm fan, all extracting. I have made several posts here, recommending that effective cooling is achieved by a design which controls the flow of air - not by simply filling the case with holes. Subjectively, I could feel that the airflow into the fanned cage was similar to that of the un-fanned cages, simply by placing my fingers in front of the perforations in the drive bay doors. This was borne out by the thermal behaviour of the drives. I suspect that this may be in an effort to achieve equal flow across the width of the cage, while not having sufficient available height to install larger fans, without having to build in ducting which would increase the depth of the cage. Sure, but my comment was in relation to the change in noise level when going from 2 fans to 6 fans. At most, the actual noise level should not be more than treble, but that is a relatively small change in a logarithmic scale.

I suspect that it wouldn't be very much louder - our hearing sensitivity is logarithmic and the sound source is in the middle of the case. That is certainly true - perhaps they could be sourced from cpu coolers, or something - but 10mm is very thin, too. For me, by far the best solution is to rely in the case fans to draw air through the drive cages. The fan connection is an in-line 2-pin layout, on a standard 0.1" spacing. The socket is attached to the drive cage, the plug attached to the fan - unlike standard fans which have the socket attached. Okay, I've added a short paragraph. However, there is nothing remarkable about installing the drives - it's just so easy.

iStarUSA BPN-350V2-SS Trayless 5in3 drive cages. When I started building my unRAID server (more than one year ago, now) I always planned that it would support up to 15 drives, using 5in3 disk cages, although I was starting with only two data drives. With that in mind, I started off with a Thermaltake V5 case, sporting 9 unobstructed 5.25" drive bays. By April this year, I had grown the system to five drives and adding further drives would require purchasing additional hardware to mount them. It was at this point that I decided to purchase my drive cages - after much online research, I decided that the specification of the iStarUSA BPN-350 met my requirements. I already had experience of a trayless drive enclosure in my Popcornhour media player, and was attracted by this feature for my unRAID server. An added bonus is that the trayless designs tend to be shallower than those which include trays. However, a major disadvantage is that I am living in the Philippines, and the only drive cages I could find available here are the SuperMicro units (at the equivalent of USD200 each). Thus, the SuperMicro would be a 'long-stop' solution, but I really wanted to try the iStar units. I found a few reviews online for the original BPN-350, but none for the V2. The reviews of the older model seemed to have three main criticisms - poor cooling, fragile plastic parts for the drive bay doors, and insecure catches on the doors. Now, I reasoned that, in a domestic environment, where I would be the only user, the last two issues could be addressed simply by exercising care. The first issue is much more critical, especially since ambient temperatures here frequently go into the low 30s celsius. However, I could see, from the pictures on the manufacturer's website, that the V2 includes air slots in the door which, according to the reviews, were lacking on the original design. A further possible disadvantage is that the iStar units use two 60cm fans for forced cooling, and these could be the source of a significant noise nuisance. However, I had already proven that, with a few simple case mods, it is possible to force all cooling air to enter the V5 cage through the disk slots. This had already shown great thermal benefit even with the standard disk mounting arrangement, lowering drive temperatures by 3 to 4 degrees celsius. I believed that it would be possible to obtain sufficient cooling even with the fans removed from the iStar cages. At the same time, I opted to add an eight port SATA interface card, selecting the Supermicro AOC-USAS2-L8i to fit in the PCI-e x16 expansion slot. In order to achieve the full 15 drive capacity, I will need to add a further port - I plan to add a two port card, at some time in the future, which will fit in one of the x1 slots. I then went through several traumas attempting to get the iStar units, which are only easily available in USA, shipped to me. Much of the frustration centered around PayPal - the preferred payment method for many online retailers. Another problem is that many iStar retailers do not hold stock, but arrange shipping directly from the factory - this also seemed to cause great problems when I tried to use a third-party purchasing/shipping agent. Eventually, Rackmount Chassis Plus, who seem to be the only retailer who holds stock, added wire transfers to their accepted payment methods. Now I was all set but I discovered that Rackmount normal shipment method was via USPS, which gave me cause for concern because the Philippine post office is not reliable (it took eight months to deliver one package I sent from UK). At my request, Rackmount obtained pricing from DHL/UPS but these were horrendously expensive (much more than shipping agents had quoted me for the same shipment method). So it was that I placed my order for three units, accepted USPS shipping and made the wire transfer. The units were shipped on July 11, with an advertised delivery time of 5-10 business days. Time passed, and I became more and more impatient. Eventually, after correspondence with Rackmount, I visited the local post office on August 10, to find that my parcel had been sitting there since July 26! For more than two weeks, no one had bothered to inform me that my delivery was available for collection! Still, the good news was that the Philippine customs hadn't shown any interest in my parcel and all I had to pay was the standard post office handling fee of approximately one US$. Anyway .. on to my review of the product : The BPN-350V2-SS is delivered in a plain white carton, measuring approximately 11" x 9" x 8" and weighing 4lbs 14.3/4ozs. It has a slot-through plastic carrying handle on the top. Inside the carton is the cage itself, wrapped in a plastic bag and supported in cellular white polyethylene(?) foam packing pieces. Also included is a small piece of paper detailing the warranty conditions (1 year, USA & Canada only). There is also a small 'zip-lock' plastic bag containing the very small and brief user manual, two keys to operate the drive bay door locks, and 12 mounting screws. No further accessories, such as cables, are included. The drive cage measures approximately 7.1/2 x 5.3/4" x 5", and weighs 4lbs 3.1/8ozs. The main frame of the BPN-350, like most other cages, is made of aluminium. Examining the drive bay doors, it is clear that most of the parts are plastic but they give the impression that you would have to be rather heavy-handed in order to cause any damage. The locking tongue is metal and engages with a metal bar, so there is no cause for concern there. I could see, immediately, why there might be criticism of the security of the door latches. The door can be closed, and remain closed, without the latch engaging. In this condition, it would be possible for the door to pop open in response to a slight nudge. When the latch is properly engaged, it takes a very definite pull on the handle in order for the door to open. I don't feel that either of the criticisms are founded - certainly not of the V2. The back panel of the cage sports two 60mm fans, 10mm thick. There are two molex-type four pin power connectors, and between them are ten SATA/SAS data connectors. Five of the connectors are yellow, and designated as SAS, the other five are black and designated as SATA. Now, I suspect that there is really no functional difference between the connections, but I've not tested my theory by testing the yellow connectors. The connectors do accept latching SATA plugs, and they do lock in securely. Looking into the empty drive bays, it can be seen that there are sizable apertures for cooling air to flow through the backplane. Installation Installation in my Thermaltake case involved: Removal of side panels. Removal of front bezel. Removal of external drive slot blanking pieces from the bezel. Removal of the machined-in drive slot blanking plates, and all installed drives/mountings. Removal of the 'tool-free' drive fixing devices. Flattening of two out of every three drive support ledges - using the largest G-clamp I could find (as advised elsewhere in this forum), although, perhaps, a 12" clamp was overdoing it! As aware as I was that only two out of three ledges needed to be flattened, I did find that I started to flatten one of the ledges that I wanted to leave in place. Now, with a half-flattened ledge, it is impossible to install a cage - I can assure you that flattening the ledges is child's play compared with trying to reinstate one which has been partially flattened! Anyway, with the ledges flattened, the cage slid in very easily, without binding or jamming - no force necessary. Now, I refitted the bezel, in order to determine how far the cage needs to be inserted to fit level with the front bezel, and which screw holes to use. In fact, the best position was obtained by using the furthest forward screw holes. No prizes, but who will be the first to say what is wrong with the photo above? When mounted like this, there is ample clearance between the back of the cage and the motherboard. I felt that it wasn't necessary to use all 12 fixing screws to fasten the cage into the case and that 8 was more than sufficient. Screw alignment was spot on. I installed the first unit intact, as intended by the manufacturer. The other two were installed with the back/top panel, including the fans, removed - a simple operation involving the removal of 12 screws and disconnection of two fan connectors. I have connected all five bays of the lowest cage to mobo ports, with the sixth mobo port connected to the e-SATA port on the top of the case. The other two units have the first four bays connected to the four connections from each port on my AOC-USAS2-L8i interface card. This leaves the fifth bay in each of the top two cages unconnected - awaiting a 2-port SATA card. I installed three drives in the top cage, and three drives in the bottom cage. I swapped the modular cables on my Seasonic PSU, from the SATA cables to the molex type cables and they connected very easily. The connectors on my PSU are the type which have 'squeeze-to-remove' grips on them, and these did not cause a problem. In use Installing drives into the cages is so very easy - just rest the drive on the door spring, line it up with the bay separator guides and slide it back gently until it contacts the ejector arm. Now, simply close the door, and the drive is pushed fully back, engaging with the electrical connections. On powering up, the noise from the fans in the top cage was very noticeable - not unpleasant, but definitely there. Whereas I normally have to be within three feet of the machine in order to hear any noise, I can now hear it from 15 feet away. Once a movie is playing, I doubt that I would be troubled by the noise but the perfectionist in me wants to eliminate it. I observed that the fan light on the top cage shows green while the lower two cages show red - not surprising really - the lower two are notifying fan failure (because I have removed them). Below each drive bay is a legend HD1 - HD5, from left to right. On the occupied bays, the legend is lit with a blue light. When the drive is in use, the blue is supplemented by a red light, giving a split red/blue indication. The legend is not lit on the unpopulated bays After a short play to gain confidence that there was no major problem playing movies, I started a parity check. After half an hour, the temperatures of the drives in the fanned cage were no different to those in the cage with fans removed - the highest temperature was only 8 celsius above ambient. The temperature of the hottest drive peaked at 39 (ambient 31), then fell back as ambient temperature dropped. Thus, I am convinced that the air flow from the case fans alone is sufficient to cool the drives, and the cage fans are unnecessary. Also, I think that it is extremely unlikely that drive temperatures will ever exceed 45 celsius. However, as a precaution I want to investigate the possibility of controlling the fan speeds according to drive temperatures. Although I wouldn't expect it to ever be necessary, in an extreme situation I would rather have a slight increase in noise than to have an overheated drive. I confidently left the non-correcting parity check running and went to bed (3am). However, during the early hours, we had a power cut and, because I had not connected the usb from the UPS before the machine was started, the UPS was unable to initiate a controlled shutdown. When power was restored, one of the data drives red-balled. I still need to investigate that fully, but I don't believe that the cages have contributed to the failure in any way. ================================================================================ Edit: It turned out that this drive failure was caused by some interaction between my SATA expansion card and the newly installed cages. The problem was cured by a firmware update to the SATA card. ================================================================================ I swapped the drive ... so easy - simply stop the array, open the bay, pull the drive out, insert new drive, close the bay door, re-assign the disk/drive and start the array. The rebuild has completed without incident. Again, the temperatures were perfectly controlled, never exceeding 40 celsius during the middle of the day. I have now removed the top/back panel, with fans, from the top cage. Conclusion Would I recommend the iStarUSA BPN-350V2-SS to others? In a domestic environment, the answer is a resounding 'YES' - in a professional environment, it might be worth training any operators to use a light hand when servicing the cages.

That Rosewill also claims "Auto Fan Speed Control".

Ooops, did I say, or imply, that? I believe that my Supermicro AOC-USAS2-L8i started life as a simple IR mode device, operating with the mpt2sas driver. I reflashed it to IT mode. Anyway, I have now flashed it with the P10 firmware which you uploaded, and everything seems to be fine.

If you have no success in finding an alternative, I could consider putting them up on my host - I have some spare capacity. Do you know how much space do they take up, and how much traffic is generated?

Indeed - it is a very simple process. Since the one thing we all have in common here is that we run slackware systems, promoting the use of the Linux versions will make life very much more simple ... no need to move cards, no need to build DOS boot usbs etc! The one thing I would advise - although I have flashed my card with the array still up and running, any access to a drive connected to the card during the flash will result in errors and possible data corruption. Much better to ensure that the array is stopped before flashing.

Sorry! I'll have a go at loading P10 firmware in a little while, then put together an archive with all the Linux versions of sas2flash that I can find.

Okay ... as you know, I'm just waiting for the combined forces of the USPS, the Philippines customs and Philippines post office to do their stuff!

I tried FreeDOS, and quite a few others, on USB stick before I gave up and flashed my card under the standard unRAID Slackware. Mine reports exactly the same - except that it's rev 02. I wonder what the difference is.

You say that your build is almost identical to the one you reference, yet you select a higher capacity psu? Why not stick to 750w as in the system you're emulating?

IIRC, the ATX2.2 standard specified a current limit for any one rail, but the 2.3 standard changed that.

It should do, yes. Hmmm ... LSI appear to have changed their product line and their website, so my link no longer works. Perhaps I should host the essential programs on my own site.

I guess that it's possible that the most worn keys might give a clue to passwords.