#Network probs
29 messages in this thread
I just got another error message while rendering on the network qeue:
A serious disk error has occurred while writing to drive D
I have gotten this message two other times since installing R3. Both
other times were while writing large (150,000+ faces) meshes to disk.
What is going on?
This looks like a SMARTDRV message. Two things that would cause this.
Its cache data got corrupted (again the memory thing) or your SCSI
bus got locked up somehow (like your ASPI driver getting corrupted).
Gee louise John, you should buy this humongous horse shoe and nail it
to your computer! <g>
The horseshoe things sounds like it'll work. I'll go visit my
ex-horse and get one off him. He's 17.2 hands (BIG), so that oughta
fix the whole computer thing in a second <g>.
P.S.: I suspected a smartdrv error, too. I'm going through my memory
stuff again, but haven't found an evildoer yet.
I seem to remember someone mentioning something on the forum about how
using smartdrv with 3DS was a very bad idea. Do you remember
anything about that?
Yeah, I remember that thread. I've disabled Smartdrv. I thought that the
newer versions would not interfere as bad, but I guess I was wrong.
I would like to be able to use it for window, becuase it speeds up operations
by several hundred percent. Perhaps someone can suggest a good batch file,
including it before booting windows, then removing it after exiting. I've
tried it a bit, but haven't got it right.
Or maybe there's a better cache that'll speed up windows but ignore 3DS.
Thanks,
John Tissavary (La Luna cie)
Although I've never ran into them myself, I've heard of
SmartDrive/Phar Lap problems quite a few times over the years. If you
really want to run a disk cache for Windows, I remember reading about
one that is highly reccomended by QuarterDeck for use with QEMM, but
can't remember what it was. I'm sure they could help you out if you
gave them a call. Maybe someone here can jump in with this…
Thanks Alan. I'll try Quarterdeck.
Unfortunately the Pharlap conflicts don't seem to stop there, either.
I'm getting error crashes a lot since installing R3 and using it on
my 3 node lantastic network. I've tried everything I can think of to
stabilize the memory conflicts, but can't seem to get it right. If
you know of anyone in the SF area that is a competent consultant and
would be able to straighten out this kind of mess, I'd be
appreciative.
At least I can get work done and rendered, even if it's not on 3 machines all
the time.
Thanks,
John Tissavary (La Luna cie)
Sorry to hear about so much trouble, and I hate to mention the
obvious, but if I were trouble-shooting your system at this point,
here is what I'd do:
1) Take ALL the systems down to straight, plain vanilla 3DS/lantastic
set-ups with NO fancy video cards (just plain VGA), NO Windows start-up
configurations or ANY other Hardware/software configurations other than
what is needed for 3DS and Lantastic to run. Run QEMM in a simple config
JUST to run 3DS and render smoothly throughout the network.
2) If it doesn't run like this, you've got real problems in your basic
hardware/software configurations and they shouldn't be hard to track
down. If it DOES run smoothly, leave it alone and get a little work done
for a few days, until you are SURE there are no problems. Run seperate
config.sys/autoexec.bat configurations for your Windows apps, or ANY
other app that doesn't like your present set-up. This is the way I ALWAYS
do it and it is the way I adivse my clients to do it also. True, it may
be possible to have one config for everything, but let's get things
WORKING first before we try to get too fancy!<g>
3) Once you are comfortable with the system's performance, add hardware to
it ONE piece at a time, carefully excluding where you need to to get
things to run right. When you get the critter(s) causing the problems,
you'll know it as SOON as you install it! If you can keep doing this
until every thing lives and plays "nicely" together, you'll have
accomplished a wonderful thing — good luck!
Please note, John, that it sounds to me like you may be trying to run
just too much stuff in too few machines with too few configurations.
Don't get me wrong, I know that ANYTHING is possible, but if you try
to keep things as SIMPLE as possible (even if it means a little
inconvenience) you just may find it may get you up and running (and
KEEP you running) a little faster and little longer.
Now, are you pulling support files off of slaves while they are
working? <key question> That can be baaaaad news.
Can (or have you) outline your FULL system and get it to me (either
posting it here or in email)
What I would like to see is a list of all hardware, brand names,
add-ons, type of network cards, display cards, config.sys,
autoexec.bats and so on, and so on. I'm not a Lantastic expert but
I'll take a look at your setup and see if I can decifer what's going
on.
I really appreciate it, Jamie. I'll E-mail you a file with all the
requested info, plus a Microsoft diagnostic session. I've begun to
stabilize the computer already, but things don't seem to be 100% yet.
I do this all the time, John. check a previous thread on this. You
can certainly launch smartdrive from a batch file just prior to the
win command in the same batch but you will need to reboot to disable
it after exiting from windows which can be also part of the batch
file. I wrote a small com file in debug a few years back that
reboots the computer when the file appears in a batch file. I've had
intermittant problems with smartdrive and 3DS rendering the nearly
consummed all my ram. It showed up as missing pixels totally black
frames in the targa files.
>> I've had intermittant problems with smartdrive and 3DS rendering the
>> nearly consummed all my ram. It showed up as missing pixels totally
>> black frames in the targa files.
PMJI (as usual!), but we have run SmartDrv and QEMM with 3DS on all
16 systems since R1 with absolutely no problems! FWIW.
I guess that proves that I've got a problem somewhere else. I wonder if my
SCSI controller could be a problem?
I'll call Adaptec tommorrow.
Thanks,
John Tissavary (La Luna cie)
Do you ever page to the HD? I recall you running 48 or 64 megs RAM on
each machine. You probably never do really use smartdrive when you
render. I'll also bet if you did see some of your machines paging
you'd say, "What's this! paging? We can't have this! Order another
64 megs for each machine!" <G>
Yeah, I page once in a while, and I did get more RAM for the main
workstation because I thought that may hinder Lantastic's
reliability. I was right, because now everything works fine, expcept
for rapid Video Post stuff (no geometry rendering, just rescaling,
transitions, etc…) done on Network Queue. That fails machines like
clockwork, and I've done so much diagnostic work with hardware and
software, with tech support, etc… that I'm beginning to suspect a
R3/Lantastic limitation here.
It's definitely not an inherent problem in R3 or Lantastic, John. As
I said, the 3DSr3 network renderering code was developed by Tom
Hudson on a 4-node Lantastic network, and it passed every test
imaginable. Plus, at least 2 dozen of our beta testers used
Lantastic networks. It's GOT to be a problem with your hardware,
bios, etc.
I'm going to switch out the HD controller as soon as I can. It's a
RISC co-processed affair, and may be the problem. My MB is an AMI
Enterprise III, and I know several people with the same MB/BOIS
version who have no problems, so I doubt it's that.
Thanks,
John Tissavary (La Luna cie)
Don:
>> Do you ever page to the HD?
We page once in a while, not often! We do have a minimum of 48 megs
on each of our 486/66s on the farm. Actually, just last night, we
hit some serious paging, not because of a big file, but because we
are creating a big video that will show off all of our "Action IPAS"
routines.
R3, just like R2, does not release the memory used by an IPAS, and
after running 20 or 25 of them the paging gets rather serious! The
only solution is to exit and re-start 3DS. The file was not that
complicated!
when you have multiple render jobs to run overnight and paging is a
problem, it might be beneficial to go back to batch rendering. At
least then, memory would be freed up between jobs. I still find a
need to use batch renders occasionally because of that reason.
Probably isn't a big problem for you, oh man of many memories, but
for others with less ram it might be a good tip.
Good tip, oh man with small (short?) memory! <g>
I've got one machine that has only 16megs ram for the time being, so I'll
definitely try batch rendering on that machine.
Have you tried Hyperdisk? I have been using it for years with all
kinds of software (QEMM, Windows, 3DS, AutoCAD, Lantastic Z, etc.)
and have had no problems. It also happens to be one of the fastest
caches you can get. It's a sort-of-shareware program available from
the Hyperdisk BBS in San Francisco. The number is (415) 882-1735.
By the way, I see from your other messages that you are looking for
some help on getting your Lantastic network rendering up and running.
Now I am _NOT_ a network "jock" but I did get my little 2 machine
Lantastic Z system up and rendering with QEMM AND a cache and I DO
live in SF. (Lantastic Z is supposed to be exactly like regular
Lantastic except it runs over the parallel ports rather than an
Ethernet card.) I can't offer expert guidance but if you can't find a
"real" guru maybe I can get together with you and between the two of
us we can hash it out.
If you're interested, contact me by E-MAIL and we'll swap phone #'s
Check your E-mail.
Thanks,
John Tissavary (La Luna cie)
John,
Checked my mail and there's nothing there. Try again?
Greg LeVasseur
I'm sorry, Greg. Check again. This time fer sher <g>!
John Tissavary (La Luna cie)
I use Smartdrv with 3DS all the time. If memory is getting hosed, it could be
causing this problem but it's not a problem I've seen.
Jamie
Autodesk, IDL
Hi Jamie!
What version of Smartdrv are you using? I had read that some versions
were causing file to become corrupted (the version supplied with
MS-DOS 6.0 I believe).
Have you tried PC-KWIK's disk cache? I've heard it's the bes on the
market, FWIW.
I'm using both DOS 5 and 6's Smartdrv and on a wide range of CPUs.
I've not seen any file corruption that I can attribute to Smartdrv.
On the other hand I have seen problems when you read/write to a
Peer-to-peer Server/Client while it's also Running 3D Studio or
AutoCAD.
IMHO – Peer-to-peer networks were never designed to handel the traffic
that a program like 3DS can generate. You can get them to do amazing
things but it's a data risk. I worked exclusively with Nwlite for
more than a year before switching to Netware 386. Nwlite is a
excellent low-end network but when it comes to heavy usage, it'll
start giving you trouble. (like corrupted files)
When this cold I'm fighting finally gets me to stay home from work,
I'll probably write my network memoirs…
This is exactly what I believe my network is suffering from. I've
bumped up some of the Lantastic startup parameters, and started up
all TSR's AFTER Lantastic (as it says to in the troubleshooting
section of the manual – a place I've neglected to search until last
night) and things seem to be okay for now. We'll see.
I've sent you the information you asked for, and am really gratefull
for any input you can give me, because I'm not sure I'm out of the
woods yet. But I am making progress with the help of people like you
and Gus and Greg.