Snap 705N/4100 lives and thrives!

[jontz]

Originally Posted by blue68f100

This thread is gone down hill, but it has been fun.

Oh come on, we can have a little fun, can't we?

Originally Posted by blue68f100

There nothing says you can't use a external power power supply. As long as the grounds are tied together for reference it should work. Put all of the drives off the external.

Buy a 8 bay drive unit w/ps then you expand the unit with a 2 drive ribbon connector. Expanding the raid 5 array up to 7-8 drives ( 1 spare). 7 x 300 = 2.1 T - overhead may be 2 tera bytes.

How sweet would this be .........

2 TB would be sweet, however external drive bays do take away from part of the beauty of a snap server, which is large, easy to configure storage in a small package. Think about it, if we could crack the LBA48 bit issue, you could have 4x750 GB drives in 1U of rack space...

[Phoenix32]

Originally Posted by jontz

Hey, if you weren't a smart ass I probably wouldn't like you

Heh, we are gonna get along great...

[Phoenix32]

Originally Posted by re3dyb0y

Joe doesn't mind

He'd filter it if he did

I kinda figured, but as an Admin of another forum myself, I figured showing proper respect was a good idea.

[Phoenix32]

Originally Posted by Parity Error

Yes, this is the right approach. Attempts to limit the starting surge will not solve the bigger issue that you really do NOT want to screw around with marginal power supplies, period. Murphy will bite you on the *ss for doing so.

Agreed. Mostly because I have believed strongly for a long time now that Mr. Murphy (of Murphy's Laws) follows me around, just to make new laws. And, he usualy tests them out on me.

[Phoenix32]

Originally Posted by blue68f100

Expanding the raid 5 array up to 7-8 drives ( 1 spare). 7 x 300 = 2.1 T - overhead may be 2 tera bytes.

How sweet would this be .........

Why go that route? There are 750GB Drives out there now I am sure of, and a friend of mine said he saw a 1TB Drive on a distributor list the other day (already for sale). So why not just go 4 x 750 or 4 x 1000 with an upgraded 1U or 2U power supply?

Originally Posted by blue68f100

This thread is gone down hill, but it has been fun.

Okay, Okay, I get the point!

So back on topic (at least part of it)...

Are we agreed that in a 4100, the power supply is capable of 4 x 160 drives, which already exceeds the LBA48 limit on them? So no problems there (yet)?

-AND-

Are we agreed, that 1) without adding in a new delay or current limiting circuit, or 2) changing the power supply in some manner, that a 4000 should handle any version of drives x 4 so long as they require no more than;

1) 2.25 Amps Spin-Up power (9 Amps Total)

AND

2) 1.125 Amps normal operating power (4.5 Amps Total)

on the 12 volt rail?


If, and I repeat IF, we are agreed, then what are the largest drives from the different manufacturers that will meet these requirements? (I am assuming we should always be good on the 5 volt rail)

The catch to this question above is finding actual requirements of a given drive versus what the manufacturers post on their web pages. Even Storage Review has very limited info in this area (the number of drives listed). I think in this case, we are gonna be back to either doing a crap load of measurments on drives we don't have, or finding a source of people who have changed drives and sort out what has and has not worked (not just drive size, but manufacturer and model number drive). Maybe, just maybe, we can do a little SWAG (Scientific Wild A#$ed Guess) and just guess based on the number of platters used in a drive versus the number of platters used in upgrades we know worked. This last one would probably preclude Seagates because they seem to draw a lot more power on the 12 volt rail than other manufacturers. Any ideas?

[Technophobe]

WRT Larger drive I have had 4 x 250Mb Quantum 2500BB drives running in a 4100 without problem. The big issue is start up current. I have noticed that my 4100 DOES NOT spin-up all the drives at the same time, but each one comes on after a short delay (app 10 secs) after the previous one. Maybe there is a BIOS tweek that could lengthen the delay as the bigger drive do take longer to come up to speed. The other thing to check is not the total power budget, but the power requirements on each supply voltage as it is likely to be the 12V supply that is pulling too much current. If so an additional 12V onlt supply may give the extra necessary to keep the supply up during start-up. Some drive can pull 2-4 Amps on the 12v supply for the first few seconds, that's a lot of (peak) power worst case over 200W plus the 5v supply of perhaps another 50W .......

[jontz]

Originally Posted by Phoenix32

Maybe, just maybe, we can do a little SWAG (Scientific Wild A#$ed Guess) and just guess based on the number of platters used in a drive versus the number of platters used in upgrades we know worked. This last one would probably preclude Seagates because they seem to draw a lot more power on the 12 volt rail than other manufacturers. Any ideas?

Seems like a three platter drive or smaller would work from what I am hearing. 4 platter drives and larger are out because of the startup current needed. I would guess that even 5 platter drives don't draw much more current than a 2 platter once they are going, but there just isn't enough umph out of that 4000 PS to get them going.

[Phoenix32]

Originally Posted by jontz

Seems like a three platter drive or smaller would work from what I am hearing. 4 platter drives and larger are out because of the startup current needed. I would guess that even 5 platter drives don't draw much more current than a 2 platter once they are going, but there just isn't enough umph out of that 4000 PS to get them going.

This is what I suspect as well... Except in the case of Seagates, I am still not sure on them because they draw lots of 12 v juice at startup compared to the others (or so it seems).

[blue68f100]

Looking at the debug cmds for a 2200, it does not give any options for delay or stager start.

Need to have someone (jontz) look at all of the debug cmds for the 4100 or someone with a 4000 to see if its a opton.

I know different version of the OS have different option, I always assume it was the same for all platforms.

[jontz]

I looked through all of the debug commands in my 4100, there is no option to either enable or disable the delay startup routine. I'll keep digging though...

[blue68f100]

I wish we had the hardware and software to use the data port on the MB.

[jontz]

Yeah, you and me both. I might have to pick up another 4100 just to mess with it. I am a little scared to take down my current box as I use it quite a bit...

[blue68f100]

Under stand that.

I purchased an flashram/bio/... programmer with the optional TSOP40 to read my 2000 flashram. Then discovered their was 2, one in socket, other soldered in.

One of these days I'm going to try to repair with the parts I picked up. If that doesn't work, I'm going to un-solder the other chip and read it. Then merge the files.

Then the fun begains.

For some reason I'm always busy. I think it because I'm so slow, due to my back.

[Phoenix32]

Originally Posted by blue68f100

Need to have someone (jontz) look at all of the debug cmds for the 4100 or someone with a 4000 to see if its a opton.

This is from my 4000 with OS 3.4.805, hope it helps...

[blue68f100]

My 2200 with 4.0.860 has more cmds than than the 4000 with 805. So there nothing there to help us.

[Phoenix32]

Originally Posted by blue68f100

My 2200 with 4.0.860 has more cmds than than the 4000 with 805. So there nothing there to help us.

When I install the 4.0860, I bet it changes to more like what you have. I will send you a copy when I do the update. I have the OS, just not installed yet. No need until a little later on (for the MS AD), so why use it yet?

[re3dyb0y]

No there is no reason

And it will only hog more memory

[Phoenix32]

Originally Posted by re3dyb0y

No there is no reason

And it will only hog more memory

Exactly!

[Technophobe]

Hi, I have an interesting problem, which maybe someone can shed some light on.

I have used both 4100 and 705N servers.

I have found the 4100 to be very relaible but I have had two 705N which appear to mess up the Raid 5 array from time to time

This generally happens when copying large numbers of (small) files from one place to another ie for mirroring or backup ..... I keep a large number of files spread across a few 4100 / 705n / 1100 servers in raid 5 where possible.

The problem is not catastophic as I can generally rebuild the raid, but this put the server off-line for some while and is annoying.

I have just loaded the latest troublsome 705 with 4 160Gb drives and all appeared to be fine , I then started to mirror another server and after about half an hour .......... Uggg

Both were running the 3.4.xx series SW (being the latest that can easily be obtained)

I assume the only real difference between the OSs is cosmetic, so Any ideas wht the Dells should appear to be more troublesome.

There a very large number of Incorrect Block Counts I've only pasted a few here for example.

The error log follows (in reverse chronological order):

File System Check : Incorrect block count I=1745664 (16 Should be 0) (Corrected) Disk 60000
RAID 5 22/6/2006 8:28:04
W File System Check : Incorrect block count I=1744896 (16 Should be 0) (Corrected) Disk 60000
RAID 5 22/6/2006 8:28:04
W File System Check : Incorrect block count I=1744128 (16 Should be 0) (Corrected) Disk 60000
RAID 5 22/6/2006 8:28:04
W File System Check : Incorrect block count I=1743360 (16 Should be 0) (Corrected)


Call Stack :
$002224EA
$0021CC66
$0021C79E
$002267F5
$00226981
$002088DC
$00208813
$0021F309
$0021A6F3
$0021FF77
$00205A41
$00205ACA
$0017610E
$0016A127
$0016A278
$00169B90
$0010413C
$001041C3
================================
System 22/6/2006 7:57:24
F PANIC : blkfree: freeing free block

Prior to this all appear to be OK.

Any thoughts would be most welcome
P.

[Technophobe]

This posting has been removed

[blue68f100]

These can be good or bad. Think postive here.

Stacks are generaly memory related, which can lead to disk corruption.
I would open the unit up and re-seat every thing, ram, ide cables, cpu ... This will cleanup all of the connections. While you have it open make sure your fans are working.

Are you running any extra demons, JVS ...
How much ram in in the unit? If you are running low on ram it cause problems too. And for some strange reason as the HD capacity goes up they like more ram.

V3 OS supports MS AD 2k but not 2k3.
4100 do not support LBA48bit addressing as you proably allready know.

These are not very fast units (as with most snaps), I'm also wondering if you may be having some network (packet) lost/collision. As networks get faster these old units can't keep up.

Just a thought.

[Technophobe]

Thanks, Actually I found a loose Ram chip (although I have had it happen on other servers)

RAM Size is 64Mb (old drives were 4x30Gb) now 4x160Gb so could be an issue.

I think my reliable servers have 128MB (but can't easily check as hidden in a remote place !!! - unless I can check through debug)

I installed a clean copy of 3.4.780 with no extras (the server had been PW'd and could not get access to it, so was the only way !!!) (Adapted just gave me access to 3.4.807 !!!)

If anyone wants it, it's here:



You say not very fast - I can get combined throughputs (ie multiple transferes)very close to 10MB/sec - I find the tranfer rate is limited by Windows not the Server, which usually puts a limit of around either 3.5Mb or 5.8Mb - no idea why (does not seem to be related to PC's CPU speed).

The 1000 / 1100 / 2200 etc series do appear to limit at app 1.6MB/sec.

Anyone know what type of RAM chips these take as Will upgrade to 128MB or more if desireable ??

LBA limit yes I know this, but going from 4x120 to 4x160 gave a worthwhile extra >40GB of storage (Raid 5)

(I did try 4x250 before I learned of the limit .... costly mistake !!!)

[blue68f100]

There is a reset procedure to clear the admin password. Do you need a copy of the admin manual and reset procedures?

What is special about 3.4.807 compaired to 805 did they say?

12.5 Meg/sec is the max for 100mbps over IP. The RAID 5 will generally max it out.

These unit use DIMM's chip PC66/100, I think some users said they needed the low profile. Issue "debug memory" from debug to get full detail on installed ram.

The over size drives should not cause any problem.

Your example is a home built NAS/serverpc running RAID 0, and SNMP is having a problem too. I would edit that post and clean it out, it really doesn't apply to the snaps.

[jontz]

Originally Posted by blue68f100

These unit use DIMM's chip PC66/100, I think some users said they needed the low profile.

Correct. Standard height DIMM's will work, but you can't get the cover back on

[jontz]

Originally Posted by Technophobe

You say not very fast - I can get combined throughputs (ie multiple transferes)very close to 10MB/sec - I find the tranfer rate is limited by Windows not the Server, which usually puts a limit of around either 3.5Mb or 5.8Mb - no idea why (does not seem to be related to PC's CPU speed).

Yeah, the 4100's will max out a 100Mb connection if they are in RAID 5 config. Not too shabby for NAS. I wish there was a way to upgrade these to Gb ethernet. The 4100's would REALLY scream then!