SnapServer M4100 - Help needed pls!

[jw2k_fr]

Hi

First post! Just found this forum after googling and it looks like a godsend!

Okay, let me start at the beginning.... I have an M4100 which has resolutely refused to behave itself in the time I have had it. I have read the post about the possible hardware problems with M4100s and can confirm that I have a revision 1 mobo - Assy# 70715048-001. It does not appear to have the solder mod between C182 and C26.

Symptoms - box will boot but does not stay online for long. Will sometimes respond to button resets from the front panel.. sometimes not.

Having whipped the lid off, it appears that the CPU fan does not spin up within the first couple of minutes of booting.

So, my questions to any willing souls would be:

1) Should the CPU fan spin from power on, or is it BIOS driven?

2) Having looked to remove the fan to test it, I have found that the CPU heatsink is firmly attached to the CPU. Has my thermal paste dried to the consistency of glue or are they attached with thermal epoxy?

3) Has anyone actually tried replicating the resistor mod to address the RAID stability issue? I can take a photo of my mobo later to confirm that it is unmodded, but don't have a camera at the moment.

4) I don't want to boot and consult the debug page until I know whether I am in danger of frying the CPU (mind you if it was going to happen, probably would have done so by now!). The box came factory configured as a 480Gb unit with 4 x Maxtor DiamondMax 9s. Out of interest, where is the boot image loaded from - flash or the HDs?

5) Has anyone tried setting the device to boot from a custom Linux image? What is the spare IDE sized connector on the Mobo for? Flashing from CD? Is it actually an IDE connector?

6) The PSU is pretty noisy, has anyone modded it to replace the fan, if so, can anyone recommend a part no?

7) Is there any benefit to upgrading the SD-RAM or onboard CPU (is that even replaceable - I can't see how to undo the socket)... yes, I am an overclocker at heart

This will hopefully only be the start of my questions... your help and suggestions are VERY much appreciated!

Have a very happy 4th of July
James

[blue68f100]

Read the FAQ's on the 4100's will answer some of your question.

The cpu fan is controlled by firmware. It only spins up when it gets HOT.
Boots from flashram.
Have not looked at the PS fan, so I can comment. But if you can open the case you can replace it.
The 4100 were shipped with 2 different ram cap 64m & 128meg. If you run JVM you need 128 for it to run smothly. PC100/133 low density Max cap 256meg.

Unless you have source, and knowledge of BIOS code. Good luck. We if rulled this dead. It may be possiable to load Linux on a a 1100 or 2200 which use a std/modified bios, but have hardware markers/checks. It's beyond my knowledge.

If you want me to look at the FB1 & FB2 areas of your MB you may email them me. The resistor values they added were 10 ohms.

If you need more info, post snap vitals by clicking on the snapAppliance logo.

[Phoenix32]

Originally Posted by jw2k_fr

I have read the post about the possible hardware problems with M4100s and can confirm that I have a revision 1 mobo - Assy# 70715048-001. It does not appear to have the solder mod between C182 and C26.

The box came factory configured as a 480Gb unit with 4 x Maxtor DiamondMax 9s.

Okay, I am going to have to assume you meant you got the unit as 480 GB unit, not factory configured that way. I could be wrong here, and that is very possible on this one, but to the best of my knowledge, none of the 480 GB units were -001 revisions -AND- all of the units above 240 GB had the modification. Again, I could be wrong, but... Something does not seem right here...

Originally Posted by jw2k_fr

I have an M4100 which has resolutely refused to behave itself in the time I have had it.

Symptoms - box will boot but does not stay online for long. Will sometimes respond to button resets from the front panel.. sometimes not.

This could be so many things, it would be hard to tell you over a forum with just this knowledge. Bad memory, bad CPU, bad board, bad power supply, bad connection, the list goes on and on...

Originally Posted by jw2k_fr

Having whipped the lid off, it appears that the CPU fan does not spin up within the first couple of minutes of booting.

The CPU fan is temp controlled, albeit at too high a temp for my tastes. There is another fan jack (on the opposite side of the memory slot) that you can plug the fan into for it to run constant if you desire.

Originally Posted by jw2k_fr

1) Should the CPU fan spin from power on, or is it BIOS driven?

See above...

Originally Posted by jw2k_fr

2) Having looked to remove the fan to test it, I have found that the CPU heatsink is firmly attached to the CPU. Has my thermal paste dried to the consistency of glue or are they attached with thermal epoxy?

The Heat Sink is permanently attached (well, without damaging the CPU).

Originally Posted by jw2k_fr

3) Has anyone actually tried replicating the resistor mod to address the RAID stability issue? I can take a photo of my mobo later to confirm that it is unmodded, but don't have a camera at the moment.

I have not done it, but I have seen several versions of it and can tell you it is not a mod for the novice. I was a certified 2M technician (micro miniature soldering repair) and can tell you this is delicate soldering. It can be done with basic equipment and at home, but not what I would recommend. You really need proper 2M equipment and qualification to be sure of not damaging anything.

Originally Posted by jw2k_fr

4) I don't want to boot and consult the debug page until I know whether I am in danger of frying the CPU (mind you if it was going to happen, probably would have done so by now!). The box came factory configured as a 480Gb unit with 4 x Maxtor DiamondMax 9s. Out of interest, where is the boot image loaded from - flash or the HDs?

See my comments above about the CPU fan and factory configuration. As for the OS, it is booted from flashram on the 4x00 units.

Originally Posted by jw2k_fr

5) Has anyone tried setting the device to boot from a custom Linux image? What is the spare IDE sized connector on the Mobo for? Flashing from CD? Is it actually an IDE connector?

This has been discussed into the ground here on this forum. Overall, it has been decided it can't be done (or rather, not by the average joe around here). In my opinion, if there is a will, there is a way. All you have to do is find out how to flash the flashram on the unit, do a couple zillion test runs modifying the OS you plan to run getting it small enough and with just the right drivers (which you may have to write yourself) to fit and run from the available flashram, convert it all to intel hex code (or whatever is required for this unit's board), and poof, you got it.

Okay, now seriously. I am not one who will willingly tell you something can't be done when it can, and I would bet it could be done, but... The fact of the matter is, it would really take someone who knew what they were doing and a lot of time (read, an engineer most likely). And, in the end, just isn't worth the mass effort it would entail. Fact is, this is a software RAID unit running on a Pentium 233MHz CPU, with a maximum of 256 MB of PC100 SDRAM, on ATA66, with a single 10/100 ethernet port. I mean come on, would it really be worth it? You could build a mini PC or 1U Server and run standard OSes on it much cheaper and easier with better results. I am not trying to be insulting, but this is akin to saying, I have this old CPU laying around, can I build a hardware RAID with it? Sure you can, with some circuit board making skills, some coding of firmware, etc etc etc... But in the end, it would be much easier and cheaper to just go buy a solid hardware RAID card. See? If you or someone else wants to take this project on, more power to you, and congrats if you succeed, but still just isn't worth it, in my opinion. But don't say I said it couldn't be done.

As for the other port, it is NOT an IDE port. I suspect it was used for some other funtion at the main board factory, but it is not another IDE port. Maybe snap-tech, if you can get him to ever answer anything without sucking up or paying him, can tell you for sure.

Originally Posted by jw2k_fr

6) The PSU is pretty noisy, has anyone modded it to replace the fan, if so, can anyone recommend a part no?

Define noisy? Is the fan going bad or do you just not like the noise it makes normally? If going bad, open it up if you can and replace the fan. If you just don't like the normal noise, then nothing you can (or should) do. There are lower RPM (thus quieter) 40mm fans, but they also push less air. Less air cooling = more heat build up. More heat build up = less life on hard drives, the electronics, and the power supply itself. These units already run warmer in a typical household environment than they should be, so let's not make it worse (these units were designed for climate controlled server environments). There are also some more expensive 40mm fans that are a bit quieter with decent air flow, but in the 40mm department, the small amount of noise difference would not be much unless you a had a bunch of these units all running.

Side note: If you do not like the noise of a SNAP 4100, DO NOT, repeat, DO NOT, buy a Guardian OS unit. They are much louder. As a general rule, most 1U units are just not going to be quiet by nature.

Originally Posted by jw2k_fr

7) Is there any benefit to upgrading the SD-RAM or onboard CPU (is that even replaceable - I can't see how to undo the socket)... yes, I am an overclocker at heart

There is some gain in performance with going to 128 MB RAM over 64 MB or 32 MB, but the gains are diminishing. 256 MB makes very little difference unless you are running large drives and using JVM a lot.

You're stuck with the CPU. A faster CPU would be nice , and the Sofware RAID and LAN would be better, but it's not going to happen. Why? Because the faster Socket 7 CPUs were all K6-2/K6-3 CPUs using a higher voltage and obviously different CPU microcode needed in the BIOS. Since this microcode is not in the BIOS and this voltage is not available, well... There you have it. One guy tried a powerleap adapter, but he was never able to see or verify any change or improvements. I suspect the board is hardwired to a specific speed. Not to mention the height issue for the CPU, HSF, and adapter in this 1U case.

[jw2k_fr]

Originally Posted by blue68f100

Read the FAQ's on the 4100's will answer some of your question.

Wilco

Originally Posted by blue68f100

The cpu fan is controlled by firmware. It only spins up when it gets HOT.
Boots from flashram.
Have not looked at the PS fan, so I can comment. But if you can open the case you can replace it.
The 4100 were shipped with 2 different ram cap 64m & 128meg. If you run JVM you need 128 for it to run smothly. PC100/133 low density Max cap 256meg.

I have 128Mb, so I won't worry further.

Originally Posted by blue68f100

Unless you have source, and knowledge of BIOS code. Good luck. We if rulled this dead. It may be possiable to load Linux on a a 1100 or 2200 which use a std/modified bios, but have hardware markers/checks. It's beyond my knowledge.

Me too by the sounds of things.

Originally Posted by blue68f100

If you want me to look at the FB1 & FB2 areas of your MB you may email them me. The resistor values they added were 10 ohms.

I can't get a large enough image without passing the minimum focal length of my camera (and it going blurry). Is this clear enough?

Originally Posted by blue68f100

If you need more info, post snap vitals by clicking on the snapAppliance logo.

Clicking on the logo:

Model Software Hardware Server # BIOS
4000 series 4.0.860 (US) 2.2.1 569482 2.4.437


Thanks for the info
J

[jw2k_fr]

Originally Posted by Phoenix32

Okay, I am going to have to assume you meant you got the unit as 480 GB unit, not factory configured that way. I could be wrong here, and that is very possible on this one, but to the best of my knowledge, none of the 480 GB units were -001 revisions -AND- all of the units above 240 GB had the modification. Again, I could be wrong, but... Something does not seem right here...

This was the reason I guessed at 480Gb factory configured - it says so on the tin


Rev1 hardware, correct?

Originally Posted by Phoenix32

This could be so many things, it would be hard to tell you over a forum with just this knowledge. Bad memory, bad CPU, bad board, bad power supply, bad connection, the list goes on and on...

Yar... I'm only a noob when it comes to SNAP servers, been an MCSE for 10 years, building PCs for longer etc etc blah blah blah. Just wondered whether there were any known propensities for failure.

Originally Posted by Phoenix32

The CPU fan is temp controlled, albeit at too high a temp for my tastes. There is another fan jack (on the opposite side of the memory slot) that you can plug the fan into for it to run constant if you desire.

Great, thanks for that. Will probably look to replace the fan for a quieter one with a slightly longer flex which will reach that far.

Originally Posted by Phoenix32

The Heat Sink is permanently attached (well, without damaging the CPU).

Okay, will not worry based on your comments below.

Originally Posted by Phoenix32

I have not done it, but I have seen several versions of it and can tell you it is not a mod for the novice. I was a certified 2M technician (micro miniature soldering repair) and can tell you this is delicate soldering. It can be done with basic equipment and at home, but not what I would recommend. You really need proper 2M equipment and qualification to be sure of not damaging anything.

Hey, I know which end of the soldering iron gets warm, but maybe I'm not up to the level to do this myself :wink:

Originally Posted by Phoenix32

This has been discussed into the ground here on this forum. Overall, it has been decided it can't be done (or rather, not by the average joe around here). In my opinion, if there is a will, there is a way. All you have to do is find out how to flash the flashram on the unit, do a couple zillion test runs modifying the OS you plan to run getting it small enough and with just the right drivers (which you may have to write yourself) to fit and run from the available flashram, convert it all to intel hex code (or whatever is required for this unit's board), and poof, you got it.

I have a kurobox (hackable version of a Linkstation) - kinda like the NSLU2 and it has given me a good introduction to Linux, but I am only really at the level of being able to follow the path others have already blazed. Having read the exploits of fellow Kuro fans trying to hack that, I know that they had the advantage of a JTAG port and knowledge of the ARM based core it runs on. I shall shelve any further designs I might have on hacking it.

Originally Posted by Phoenix32

Okay, now seriously. I am not one who will willingly tell you something can't be done when it can, and I would bet it could be done, but... The fact of the matter is, it would really take someone who knew what they were doing and a lot of time (read, an engineer most likely). And, in the end, just isn't worth the mass effort it would entail. Fact is, this is a software RAID unit running on a Pentium 233MHz CPU, with a maximum of 256 MB of PC100 SDRAM, on ATA66, with a single 10/100 ethernet port. I mean come on, would it really be worth it?

No!

Originally Posted by Phoenix32

You could build a mini PC or 1U Server and run standard OSes on it much cheaper and easier with better results. I am not trying to be insulting, but this is akin to saying, I have this old CPU laying around, can I build a hardware RAID with it? Sure you can, with some circuit board making skills, some coding of firmware, etc etc etc... But in the end, it would be much easier and cheaper to just go buy a solid hardware RAID card. See? If you or someone else wants to take this project on, more power to you, and congrats if you succeed, but still just isn't worth it, in my opinion. But don't say I said it couldn't be done.

Message received loud and clear

Originally Posted by Phoenix32

As for the other port, it is NOT an IDE port. I suspect it was used for some other funtion at the main board factory, but it is not another IDE port. Maybe snap-tech, if you can get him to ever answer anything without sucking up or paying him, can tell you for sure.

Never mind... see above

Originally Posted by Phoenix32

Define noisy?

Not suitable for residential use....

Originally Posted by Phoenix32

Is the fan going bad or do you just not like the noise it makes normally?

Not going bad, just too noisy for my study

Originally Posted by Phoenix32

If you just don't like the normal noise, then nothing you can (or should) do. There are lower RPM (thus quieter) 40mm fans, *snip*

Given that the fan isn't running normally, I figure that a quieter, lower flow fan running permanently is acceptable. Before the fan kicks in, the heatsink is almost too hot to touch. With the current fan running, the heatsink is below body temp. So a slower speed (quieter) fan running permanently should strike a good balance between the two (imho). I shall investigate...

Originally Posted by Phoenix32

Side note: If you do not like the noise of a SNAP 4100, DO NOT, repeat, DO NOT, buy a Guardian OS unit. They are much louder. As a general rule, most 1U units are just not going to be quiet by nature.

Okay, good to know - thanks for the pointer.

Originally Posted by Phoenix32

There is some gain in performance with going to 128 MB RAM over 64 MB or 32 MB, but the gains are diminishing. 256 MB makes very little difference unless you are running large drives and using JVM a lot.

I may try and snag one if I see one cheap given the size of the array, but it is not going to be a high use item, so I won't worry if I don't.

Originally Posted by Phoenix32

You're stuck with the CPU. A faster CPU would be nice , and the Sofware RAID and LAN would be better, but it's not going to happen. Why? Because the faster Socket 7 CPUs were all K6-2/K6-3 CPUs using a higher voltage and obviously different CPU microcode needed in the BIOS. Since this microcode is not in the BIOS and this voltage is not available, well... There you have it. One guy tried a powerleap adapter, but he was never able to see or verify any change or improvements. I suspect the board is hardwired to a specific speed. Not to mention the height issue for the CPU, HSF, and adapter in this 1U case.

I shall stick with what I have, so long as I can get it to stay up longer than a tart's knickers

When connecting to the debug page:

Product: 2 - "Laser/Spinnaker: 4 drive IDE" - Platform Bytes "2.2.1"
Server-4.0.860, Build Date: Mar 2 2005 17:50:35
This is the NON-Debug executable, and was built by KEVIN
You are logged in as user ROOT (local)


Thanks
J

[Phoenix32]

Originally Posted by jw2k_fr

This was the reason I guessed at 480Gb factory configured - it says so on the tin

Yup, it is a 480 GB unit...

Originally Posted by jw2k_fr

Rev1 hardware, correct?

I no longer have my 4100 units to check, and my memory is failing these days, but that may be a different part number than the early ones were. And, when I was talking about -001 revision units, I was talking about model number on the rear sticker. Which incidently, does not even have a -00x number on it (and is a different model number than my 4100's were). Just to be 100% sure, check the sticker on the inside of the case and make sure the main board matches with what is on the sticker (serial number). This way you will know for sure someone did not swap boards on you etc...

Originally Posted by jw2k_fr

Yar... I'm only a noob when it comes to SNAP servers, been an MCSE for 10 years, building PCs for longer etc etc blah blah blah. Just wondered whether there were any known propensities for failure.

Understood, but not what you asked...

Take the drives out, check them with spinrite. Swap the memory out with a different stick. Hard reset the unit (4 led reset). Check the voltages on the PSU as you power the unit up. Get the CPU cooled off good (and keep it that way for testing). Do these one at a time and you may find your culprit.

Originally Posted by jw2k_fr

Great, thanks for that. Will probably look to replace the fan for a quieter one with a slightly longer flex which will reach that far.

Not a bad idea...

Originally Posted by jw2k_fr

Hey, I know which end of the soldering iron gets warm, but maybe I'm not up to the level to do this myself :wink:

Another good plan...

Originally Posted by jw2k_fr

I have a kurobox (hackable version of a Linkstation) - kinda like the NSLU2 and it has given me a good introduction to Linux, but I am only really at the level of being able to follow the path others have already blazed. Having read the exploits of fellow Kuro fans trying to hack that, I know that they had the advantage of a JTAG port and knowledge of the ARM based core it runs on. I shall shelve any further designs I might have on hacking it.

To quote and old Queen song, "And another one bites the dust".

Originally Posted by jw2k_fr

No!


Message received loud and clear

'Bout time someone got it...

Originally Posted by jw2k_fr

Not suitable for residential use....



Not going bad, just too noisy for my study

Get yourself a SNAP 4000...

Originally Posted by jw2k_fr

Given that the fan isn't running normally, I figure that a quieter, lower flow fan running permanently is acceptable. Before the fan kicks in, the heatsink is almost too hot to touch. With the current fan running, the heatsink is below body temp. So a slower speed (quieter) fan running permanently should strike a good balance between the two (imho). I shall investigate...

Now wait one dang second here! In your original question (#6), to which my reply was directed, you said PSU fan, not CPU fan. Now you are talking CPU fan being the one too noisy.

With that said, I had 3 SNAP 4100 units and I had to listen close to even hear the CPU fan over the PSU fan in any of them. If you can hear your CPU fan that clearly, then it is too loud and needs replaced (or your PSU unit fan is giving/has gone out. And yes, a quieter fan, always running, should do the job for the CPU.

Originally Posted by jw2k_fr

I shall stick with what I have, so long as I can get it to stay up longer than a tart's knickers

See my comment above for trying to correct this problem (a place to start).

[jw2k_fr]

[quote=Phoenix32]Yup, it is a 480 GB unit...

I no longer have my 4100 units to check, and my memory is failing these days, but that may be a different part number than the early ones were. And, when I was talking about -001 revision units, I was talking about model number on the rear sticker. Which incidently, does not even have a -00x number on it (and is a different model number than my 4100's were). Just to be 100% sure, check the sticker on the inside of the case and make sure the main board matches with what is on the sticker (serial number). This way you will know for sure someone did not swap boards on you etc...

Originally Posted by Phoenix32

Now wait one dang second here! In your original question (#6), to which my reply was directed, you said PSU fan, not CPU fan. Now you are talking CPU fan being the one too noisy.

Errr, yes I did Just keeping you on your toes *ahem*
I don't think either fan is failing, they are just very noisy. The CPU fan spins VERY smoothly - bearings seem in excellent nick.

Originally Posted by Phoenix32

With that said, I had 3 SNAP 4100 units and I had to listen close to even hear the CPU fan over the PSU fan in any of them. If you can hear your CPU fan that clearly, then it is too loud and needs replaced (or your PSU unit fan is giving/has gone out. And yes, a quieter fan, always running, should do the job for the CPU.

CPU fan is audible because I have the lid off for photos etc. It also seems to be at a slightly different pitch so it is apparent when it spins up. Putting the lid on muffles it greatly. Either way, I think the PSU fan would be the biggest focus of any attempts to reduce noise

Thanks
J

[jw2k_fr]

Originally Posted by Phoenix32

Yup, it is a 480 GB unit...

I no longer have my 4100 units to check, and my memory is failing these days, but that may be a different part number than the early ones were. And, when I was talking about -001 revision units, I was talking about model number on the rear sticker. Which incidently, does not even have a -00x number on it (and is a different model number than my 4100's were). Just to be 100% sure, check the sticker on the inside of the case and make sure the main board matches with what is on the sticker (serial number). This way you will know for sure someone did not swap boards on you etc...

There are stickers on the mainboard that have been stuck over the original screen printing on the board. The model number seems to match that mentioned in Blue's 'Attention all 4100 owners' post as being an unmodded unit. I shall see how stability fares when swapping out some of the components

Originally Posted by Phoenix32

Now wait one dang second here! In your original question (#6), to which my reply was directed, you said PSU fan, not CPU fan. Now you are talking CPU fan being the one too noisy.

Errr, yes I did Just keeping you on your toes *ahem*
I don't think either fan is failing, they are just very noisy. The CPU fan spins VERY smoothly - bearings seem in excellent nick.

Originally Posted by Phoenix32

With that said, I had 3 SNAP 4100 units and I had to listen close to even hear the CPU fan over the PSU fan in any of them. If you can hear your CPU fan that clearly, then it is too loud and needs replaced (or your PSU unit fan is giving/has gone out. And yes, a quieter fan, always running, should do the job for the CPU.

CPU fan is audible because I have the lid off for photos etc. It also seems to be at a slightly different pitch so it is apparent when it spins up. Putting the lid on muffles it greatly. Either way, I think the PSU fan would be the biggest focus of any attempts to reduce noise

Thanks
J

[blue68f100]

jw2k, Take another picture and backup so the picture is clear. Shoot at 2.1 mp, then I can enlarge it and get a good look. My eyes are not as good as they use to be. Send it to me via email. Too large to post on the threads.