EMC KTN-STL3 15 bay chassis

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I just picked up 5 of these. I checked the power supplies, and I have 3x3rd Gen VE ands 2x3rd Gen.
I tested them, empty but with both cords plugged in at idle:
3rd Gen VE: 40-41watts
3rd Gen: 34-35 watts

...so at least for my experience, the VE were NOT the best of the gens. Noise was about equal.
 
all power supplies have poor efficiency at low load and 5 watts is hardly more than measurement error. I'd be interested in a full load comparison though. 3 of mine with 45 6 platter drives comes to 630 watts under an 80% write load with non VE 3rd gens. that amounts to 14W per drive including the chassis. THe data sheet nominal working power draw is 11.86w per drive, meaning draw of the chassis under load falls from your 35 watts to 32 watts and that's even with a higher fan draw than an empty chassis
 
Last edited:
  • Like
Reactions: nexox
I'm not sure how anyone could accurately answer about all future possibilities, but all indications are that with the correct interposers, these shelves can run any size SATA or SAS drive. I have a couple 16TB's in mine, mixed in with 8's, 10's, 12's, and 14's, both SAS and SATA drives in the mix now too. They all run just fine.

Low negotiated link speeds when a shelf is full or near full of drives appears to be the only persistent annoyance. That is 100% solvable when running all SATA drives, but it is a somewhat cryptic fix. However with a mix of drives the problem is so far unsolved (for me, anyway) when running a SAS/SATA mix. I have not tried all SAS drives in a shelf yet, but it's probably same as all SATA (link speed can be successfully dictated). These shelves seem to run faster/better as 12-drive shelves then 15-drive, with four drives and a blank in each of the three 5-wide bays. I have not run benchmarks to truly know that, it's a subjective observation.

These units are fairly old SAS2 tech though, on the slower (and cheaper) side of what is out there. They will not be "good" forever. Which is true of anything. I would not expect to still be running these shelves ten years from now. But maybe.

I've tried to find this cryptic fix in the thread and came up empty, could you share it or point me in the right direction? I have two of these, currently 15 drives in one and 14 in the other, all SATA, and the negotiated speed seems to jump all over the place. I manually unmounted all of the 3G negotiated drives, ejected them, reseated them, and finally got everything at 6G. Then I had to shut down my server to disassemble and clean out some dust, and almost a third of them popped back up at 3G again. This does seem to cause a bottleneck when I perform activities that hit every drive, like when performing a SnapRAID sync, so I would love to know the secret to getting them sticky at 6G.
 
I've tried to find this cryptic fix in the thread and came up empty, could you share it or point me in the right direction? I have two of these, currently 15 drives in one and 14 in the other, all SATA, and the negotiated speed seems to jump all over the place. I manually unmounted all of the 3G negotiated drives, ejected them, reseated them, and finally got everything at 6G. Then I had to shut down my server to disassemble and clean out some dust, and almost a third of them popped back up at 3G again. This does seem to cause a bottleneck when I perform activities that hit every drive, like when performing a SnapRAID sync, so I would love to know the secret to getting them sticky at 6G.

You have two paths to this and both are a royal pain in the ass:

1. Get Avago MSM working for your system to find out which drives negotiate slow speeds. Strictly follow their recommendations in the readme for the correct version of JRE and put folders and files in the standard places, and modify the path correctly too (if on Windows). Else it will not work. This is tedious to get right. All this does is tell you which drives are slow. All other methods I tried will lie to you about spoeeds. MSM is the only one that won't lie. One 1.5 speed drive will slow down the whole show.

2a. Get some version of LSUItils working. This is harder, and the real trick. I don't even have that myself right now, but I have pulled off the trick in the past (when I wrote what you quoted, I had it). I wish I had taken better notes for how I got there, since after I rebuilt my server and I cannot get it to work again. My issue may be a SuperMicro board issue, they are very finicky. The Linux version of LSIUtils is easy on (I think) any board, while the Windows version is fussy-- difficult or impossible to get running for a SAS2 HBA. But with LSIUtils working, you can then simply change the speeds on a hot drive on the fly, via a menu.

2b. Boot with all drives pulled and one by one, with maybe a minute in between, add your drives. I am running a mixed batch of SAS and SATA with mixed sizes, 36 total, 12 per shelf (three shelves). It is fast, but only after I spend the 36 plus minutes booting the thing. Good Fing thing that I only need to do this about once or twice per year (am running Windows LTSC, currently). You will need to have some plan to get whatever drive management software you run there to pause, and thus not freak out on you (or worse: scramble all your shit) when you first boot up with missing drives.

The other tricks to speed with these shelves is to run all HGST SAS drives in one shelf (or all shelves) since an all-SAS shelf will all always boot all at 6gb speeds, without fail. When they are not mixed with sata on the same shelf, that is. Not an option for me (I'm not chucking 100+ TB in working SATA drives, especially not right now, with those prices).

The other trick is run 12 drives per shelf, max, in three groups of four per shelf. Big speed gains here, bigger than 1.5 to 6.0 gives. Leave the 5th 10th and 15th slots open, and then wonder why you didn't try that easy mod years ago. It is that much faster. You'll need a third shelf with 29 drives, as I ended up doing too. Be sure to fully air-block the 5/10/15 slots with a dummy drive though, else all the air goes there and your fans will go apeshit to compensate. I posted about how to do this before.

If you manage to spoof them and run the fans much slower (I have) then you *really* need to block open slots, or you will get hot drives. I think this may only apply to me, since I think nobody else has managed the quiet fan trick. Don't ask (it's too much to type) but it involves high Amp manual pwm fan controllers, mining fan spoofers, and some miocrosoldering in the EMC PSU's. If you have no crytomining mods experience then don't even try it. It's hard, and it's too easy to hurt yourself and/or hurt your equipment. But it is pretty nice. I have one big dial with an LCD under it that reads 001 to 100 (percent fan speed) and with that pwm generator I manually adjust the fan speed for all 12 fans at once. (two fans per PSU, two PSUs per shelf, 3 shelves... equal 12 formerly screaming fans in there). Do the math on the amps of 12 of these fans. It's a lot of current to modulate. I can also unplug one PSU and still keep the fans at very low speeds, like maybe at 25% or so. But since booting is such a pain, just to get 6.0 on my annoying SATA drives, I keep all the PSUs running. The high power use before this mod was all about those damn fans. My fans are all running at 19% and the whole rack is whisper quiet. But this is a big tease, since I don't recommend trying this unless very handy. If you ARE very handy, I just gave enough info about how to get there.

If you are not being kicked down to 1.5 then I'm not sure I would bother messing with 3.0 since your spinners cannot run faster than 3.0 anyway, except for their caches.
 
Last edited:
  • Like
Reactions: notsonewbie
Oh: the stick 6.0 trick: it was a batch file script for Avago MSM that I made and set up to run at boot. Which was easy enough to write but without LSIUtils running, you cannot get there.

In this age of AI, I found no help there, in getting LSIUtils to run on my new server. All that is is a system that reads crap of the internet and parrot it back while pretending to be a person. I wish it was more, but for now it doesn't really *know* anything. I think my issue is my unicorn motherboard.

Sticking wth Windows and that SuperMicro unicorn board (to get a Xeon CPU with QuickSync) made sense for me, with blue iris and plex server box all-in-one. I didn't anticipate losing the utility of LSIUtils.

After switching to Windows LTSC I very rarely need to reboot anything anymore, I highly recommend that. The pain in my ass boot procedure has therefore been reduced in frequency so much that losing LSIUtils is not a big deal to me now. I might even be able to get it running again, since (as I said) I failed to take good notes, on that. Actually more likely: I took great notes but now, I cannot find them again.
 
Last edited:
Oh: the stick 6.0 trick: it was a batch file script for Avago MSM that I made and set up to run at boot. Which was easy enough to write but without LSIUtils running, you cannot get there.

In this age of AI, I found no help there, in getting LSIUtils to run on my new server. All that is is a system that reads crap of the internet and parrot it back while pretending to person. I wish it was more but for now, it doesn't really *know* anything. I think my issue is my unicorn motherboard. Which made sense with blue iris and plex server box all in one, but I didn't anticipate losing the utility of LSIUtils.

After switching to Windows LTSC I very rarely need to reboot anything anymore, I highly recommend that. The pain in my ass boot procedure has therefore been reduced in frequency so much that losing LSIUtils is not a big deal to me now. I might even be able to get it running again, since (as I said) I failed to take good notes, on that. Actually more likely: I took great notes but now, I cannot find them again.

Thanks for pointing me in the right direction. Sounds like I have my work cut out for me.

Just a note-the newer LSI Storage authority will let you see negotiated links, and is easier to get working than MSM.
View attachment 50837

Thanks for the tip - I have to refer back to my notes because I was able to see the negotiation rates somehow without installing any special utilities on Debian-based OpenMediaVault.
 
Just a note-the newer LSI Storage authority will let you see negotiated links, and is easier to get working than MSM.
View attachment 50837

Sexy. I tried that too, but I could not get it to recognize my older HBA cards (I bought several LSI 9206-16e cards, and I like them). This may also be a SuperMicro + Window + Xeon W1290-P CPU issue. Too bad LSI Storage will not allow you to change the speeds on a hot drive. I think that particular capability is reserved for LSIUtils (?)
 
Thanks for pointing me in the right direction. Sounds like I have my work cut out for me.



Thanks for the tip - I have to refer back to my notes because I was able to see the negotiation rates somehow without installing any special utilities on Debian-based OpenMediaVault.

I think all these things may be better at "just working" when you are using a Linux-based OS. If I didn't need WIndows for Blue Iris, I would make the leap to Linux immediatly. I will, soon.

I procured a Dell PowerEdge R730xd before the price doubling ( Amazon.com ) that brings another 12 drive caddies to the show. I plan to use that to transition all my data in large chunks to a Linux-based ZFS, but it's only been gathering dust for a while. I'll eventually get around to setting it all up, after which these current Windows crap related issue of mine should all vaporize. To then be replaced with a whole new set of unknown issues (hence the dust)
 
Just a note-the newer LSI Storage authority will let you see negotiated links, and is easier to get working than MSM.
View attachment 50837

Hey, about this: does the LSI Storage Authority allow you to change the link speeds (such as LSIUtils allows) ?

If not, then I think the ultimate answer to this link speed problem is to get LSIUtils working. And then maybe add a script, like I once had in my rig.

That server got killed by lightning, sadly. Along with the magical Microcrap Windows-friendly version of LSIUtils that I once had. I can't believe I didn't keep that someplace else, it was very hard to find. If someone knows where to find it then please do share
 
Hey, about this: does the LSI Storage Authority allow you to change the link speeds (such as LSIUtils allows) ?

If not, then I think the ultimate answer to this link speed problem is to get LSIUtils working. And then maybe add a script, like I once had in my rig.

That server got killed by lightning, sadly. Along with the magical Microcrap Windows-friendly version of LSIUtils that I once had. I can't believe I didn't keep that someplace else, it was very hard to find. If someone knows where to find it then please do share
Not that I can find. MSM does, if you can get it working.
I do have MSM 17.05.01.02 running on server 2025, fwiw.
Coming back to this, as it was a bit of a pain to get this version working:
1: Use this 32-bit JVM: Archive | AdoptOpenJDK I specifically used : ibm-semeru-open-jre_x86-32_windows_8u432b06_openj9-0.48.0
2: Make sure you install the JVM to a path without spaces, ie c:\jvm8 or similar
3: you do not need the java beans activation file.
4: make the JRE_Home variable without a \ at the end, ie c:\jvm8 in above example.
Do the above, and it will work on server 2025, although discovery doesn't seem to be working to find other instances you can just enter the IP.
 
Last edited: