PDA

View Full Version : Burly SCSI ID connections



Taffy.C
09-29-2001, 05:03 PM
Hi There,

I am building a 4 drive array in a burly enclosure using X15 36g Cheetahs. I want to use the SCSI ID selectors on the back of the enclosure to set the SCSI IDs on the drives. Question is what pins go where on the drive.

I have the following connector plugs for connection to each drive:

One with black lead on one pin only
One with yellow on one pin and pale brown on the other
One with red on one pin and dark brown on the other

I want to connect these to the J5 Connector block on each drive. I have been following the "Roll Your Own Raid" instructions up 'til now (which have been wonderful), but the instructions are a little vague on this point.

Can you please help me with understanding what colors connect to what pins on the J5 block??

I am using Seagate ST336752LW drives.

Thanks .... Taffy.C

lasvegas
09-29-2001, 09:43 PM
Sounds like your connectors follow standard color coding. In this case they would be:

black = common
brown = Address 0 (dark brown?)
red = Address 1
orange = Address 2 (pale brown?)
yellow = Address 3

If you've determined the correct polarity of your LED wire, put the black lead on the same side of the connector that the black lead for the LED in on. Connect the other two to other side of the connector to follow the above pattern.

kaye
09-29-2001, 11:38 PM
Taffy,

The plug with red wire should also have a blue wire. This plug goes horizontally, red on the upper A0 pin, blue on the upper A1 pin.

The single black wire connector goes on the ground pin under AO red wire.

The other 2-wire connector, again installs horizontally on the upper row of pins, yellow wire to A2, brown to A3.

Hope this makes sense. IMO, this is the hardest part of wiring a Burly. You get this right, you are an Ace of the Base. Luck. k

lasvegas
09-30-2001, 12:44 AM
kaye's the man. Listen to him. Just ignore my post. The colors don't work that way in this case... http://macgurus.com/infopop/emoticons/icon_smile.gif

Taffy.C
10-02-2001, 11:07 AM
OK - So I built the array. Just for info, the SCSI ID selector cables are a different color in the burly I received. The colors I received are brown, red, yellow, pale brown (beige?), and black. In order to get the SCSI ID selectors to work correctly for all 15 IDs I had to set the connections as follows:

A0 - brown
A1 - red
A2 - yellow
A3 - beige

black goes on the pin under A0 as kaye said.

So, now i have a fully functional 2-channel 4 x 36G X15 Cheetah array with all leds and SCSI ID selectors working correctly. Is there some official guru graduation ceremony now? - do i get an autographed screwdriver or something http://macgurus.com/infopop/emoticons/icon_smile.gif

here are the performance figures from the array. It is currently striped with expressraid and i am running the external SCSI cables which came with the Atto UL3D kit. I wasn't sure what max file size to choose so i did a sample

512K
PR 196 SR 183 PW 164 SW 155

2M
PR 201 SR 146 PW 194 SW 168

2M
PR 216 SR 194 PW 197 SW 190

Sample size was 8 in all cases. Do these figures look like what I should expect for this array?? Thre were some curious peaks and troughs in the graphs at times - is that to be expected??

So, I have one last problem - after spending about 8 hours building the whole thing, I have discovered that my burly has a flaw - the power switch sticks and sometimes after turning it on, I can't turn it off. Can anyone suggest any solution before I order and advance replacement and change out the burly??

Thanks ..... taffy.c

magician
10-02-2001, 02:39 PM
taff

excellent work!

for future reference, here are some comments on wiring the ID connectors:

1. the MAPxxx3 enclosures (where the last numeral in the part number is a "3.") are wide enclosures, meaning they include switches which accommodate 15 addresses, used by so-called "wide" 16-bit SCSI devices, including LVD devices.

2. enclosures with part numbers ending in "1" are narrow enclosures, with switches that support only seven devices.

3. in all cases, ID 7 is never used. It is reserved for the SCS host by default. No need to set anything. Your SCSI host, whether it is a logic board or a SCSI card, will automatically claim this address.

4. a convention used here is to set external LVD devices beginning with the high numbers starting with ID 15. Internal devices are set beginning with the low numbers, starting with ID 0.

5. wide MAP enclosures with 15-ID switches typically have five colored wires coming off them inside the enclosure. If you look at the ID switches as they rest in the backplane of the enclosure, they have gold leads which are populated with the actual wires. These wires vary in color, but are typically red/brown, yellow/grey, with the pairs terminating in a black dual-pin connector. Color combinations may vary. The last color, always, is black.

6. the colored wires correspond to address pins on the rear of drives. The way to understand the sequencing is to simply look at the switch on the backplane, and note the order of the wires, from bottom to top. The bottommost wire, whatever color it is, goes on the 0-pin on the rear of the drive. The 0-pin on the rear of the drive is always the farthest pin on the top row to the right. We're talking exclusively about Seagate Cheetahs and Barracudas here. Any other drives will probably have different layouts.

7. the other wires, from bottom to top, go on the rear of the drive from right to left, on the top row. In other words, you look at the rear of your Cheetah or Cuda, and focus on the top row of pins, from far right to left. The 0-wire goes on the 0-pin, and the 1-wire (next up from the bottom), goes on the 1-pin, second from the far right, and onwards.

8. the black ground, which is always the top wire, goes on the second pin on the lower row from the left.

once you have your LED wires and ID wires attached and have confirmed that they are correct (by actually using the machine and scanning your bus w SCSI Probe or PeripheralView (both on our FTP site)), tack them down w a bit of hot glue. This will keep them in place no matter what.

http://macgurus.com/infopop/emoticons/icon_smile.gif

http://www.macgurus.com/ubb/magician2.gif

kaye
10-02-2001, 02:53 PM
Well as far as I'm concerned you get Ace of the Base status, especially since the ID wires are colored differently. Well done! Which computer is the Burly attached to? OS version?

Your numbers look great. Run the EPT test at 8MB, Sample Size 2, No Disk Cache. These are G2 36GB X15's?

The sticky Power Switch, here is what I would do before the hassle of a return/rebuild. Pull the power cord, squirt a little TV tuner cleaner in the seam around the button and work the switch many times in/out until it feels like it is working correctly. Wait several hours for the stuff to dry out before using the Burly. If that does not do it, you will need at least a replacement switch. k

Taffy.C
10-02-2001, 04:21 PM
kaye & magician,
thanks for the thumbs up. kaye - yeah they are G2 36G X15s. its running connected to a G4/733 with 1G RAM and Digital Voodoo Video capture card. I'm going to be using the system for uncompressed 10-bit video capture/playback with Final Cut Pro and Adobe After effects. Digital voodoo told me that in order to do real time dissolves I needed to have at least 4 x X15 cheetahs running on 2 channels. I guess thats because its pulling 2 separate 27MB streams of video off the array simultaneously to do that.

One question - where can I get TV Tuner fluid??

...taffy

Taffy.C
10-02-2001, 06:18 PM
P.S. OS 9.1

kaye
10-02-2001, 11:32 PM
Spray cans at Radio Shack.

I forgot to answer your question about peaks and troughs. Yes, you will see that, because, without System Disk Cache, you are looking at raw performance of the drives and their onboard cache. If you turn on System Disk Cache in the test, which is the way you would normally run anyway, the graphs smooth, dampen like a shock absorber. k

magician
10-03-2001, 03:27 PM
oops, forgot that question on the sticky switch.

tv tuner cleaner will work, or any silicon-based spray lube. I would just pull the screws holding the bezel in place (you will see them behind the front frame) and drop the bezel down to expose the plastic button itself and the metal switch behind it. (It goes without saying that everything should be powered down and disconnected while working).

you can check the functionality of the metal switch, and hit it w your lube (even a little WD40 will work--just be very precise and targeted and use a small straw: no need to glob it around, this isn't a shotgun we're working with here) if necessary, and then shave any plastic that might be rubbing on the button itself. Your lube there should slick things up a bit and take care of the problem.

if this don't fix it, let us know, and we'll swap the whole shebang out for you. That will be easier, believe it or not, than swapping out just the switch assembly itself.

http://macgurus.com/infopop/emoticons/icon_smile.gif


http://www.macgurus.com/ubb/magician2.gif

Taffy.C
10-03-2001, 08:06 PM
Seems to be just a problem with the plastic catching I think. The metal switch seems to be problem free. So, I will try just lubricating the plastic.

A couple of other questions. When I power down my mac with the external array still powered on, sometimes I get drive activity leds coming on and staying on until I shut down the array. It sometimes does not happen at all, and when it does happen it seems to be on one or the other SCSI bus - i.e. both drives on the bus are affected. Does this sound normal?? I have not seen this before (but then I've never had an external SCSI array before either!)

Another question - what does it really mean when you benchmark a drive to say 200MB read and write speeds like I have? Is that the best case running on the outer edge of the platters?? Should I expect significant degradation in speed of the array if the inner sections of the platter are being accessed.

This is important because I may want to use this array for some 10 bit high def video work which requires sustained worst-case performance of approx 170 MB/S. How do you benchmark for worst case??

taffy.c

kaye
10-04-2001, 10:36 AM
My Burly, I have never seen the LEDs come on and stay on until I shut it down. Yes shutdown the Mac first, then I see momentary LEDs come on and go off, then I shutdown the Burly. Maybe something with ExpressRAID but I have not seen it with SoftRAID. Could be that ExpressRAID is trying to tidy up the drives but can't because the Mac is shutdown but that is a guess.

Yes, that testing is the best case running on the outer edge of the platters since you have not partitioned the Striped RAID.

Since you have not installed any software/files on the RAID, you could repartition the drives into several partitions, each with a unique name such as Burly.1, Burly.2, etc. and then test each partition. The last partition installed would be the slowest, on the inner part of the platters.

Your worst case senceario needing 170MB/s, how big are the file sizes? Pretty steep requirement. k

Taffy.C
10-05-2001, 05:47 PM
Mmmm - you were correct Kaye - when I switched to softraid, I don't see the drive leds staying on on the array anymore at shutdown - another reason to use softraid! http://macgurus.com/infopop/emoticons/icon_smile.gif

.... taffy.c

Taffy.C
10-06-2001, 11:34 AM
Kaye - I tried partitioning into 4 partitions like you suggested to try to find worst-case performance on the inner sections of the platters. Each partition is striped across the 4 drives. But they each have roughly the same performance. Either I am doing something wrong or seagate have figured out some way to defy the laws of physics http://macgurus.com/infopop/emoticons/icon_biggrin.gif Any ideas??

TZ
10-06-2001, 12:33 PM
The review of the latest X15's on www.StorageReview.com (http://www.StorageReview.com) pointed out how Seagate managed to maintain performance better on inner tracks. I'd like to throw 4 Atlas 10k III's just released that are showing 56MB/s SW. Your setup seems to say that there is little degredation - each drive adding nearly 50MB/s.

Gregory

kaye
10-06-2001, 08:20 PM
I never saw an appreciable loss on the inner partitions either. But then I didn't have a minimum need of 170MB/s either. k

Taffy.C
10-10-2001, 04:54 PM
Wow - I just became a real member member - no junior member anymore! At this rate I should get to be a Guru Forum Ruler in about 2120 - next life maybe.