This is topic Networking Woes in forum Film-Yak at Film-Tech Forum ARCHIVE.


To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=8;t=001001

Posted by Mark Gulbrandsen (Member # 72) on 06-25-2002, 12:15 AM:
 
Just finished setting up a new Box Office For Windows ticketing system here in Aspen, CO. Am having terminals dropping off the network at random. Seems to happen more often when the system is really busy, but this also happens when a terminal(s)is idle.This can happen at any time and re-booting the computer brings it back on line just fine. We have several other installations of this system (Box Office For Windows 32 Bit) running in other locations just fine. Any suggestions as to why this sort of problem can occur on a system would be appreciated. The OS here is XP Pro, Linksys 16 port Switch, All terminals networked back to a main ticketing system server built around Celeron 1.2 ghx cpu's, 256 mb ram.
Thanks in advance for any suggestions and input.
Mark @ GTS
 
Posted by Daryl C. W. O'Shea (Member # 1303) on 06-25-2002, 12:34 AM:
 
Mark,

Sounds like it may be a routing problem with the switch, a little extra info about the network may help...

Have you assigned IP address and subnet masks to clients?
What operating system are you running on the server?
Do the clients logon to the server?
What make/model are the network cards?
What network protocol are you using? Is it the only one installed?
Can other stations see the dropped stations?


Daryl
 


Posted by Steven Gorsky (Member # 603) on 06-25-2002, 12:54 AM:
 
Daryl's list is a great start, but I have some additional questions.

Were all network cables tested after being installed?
How many terminals/clients?
What brand are the computers? (There are some systems and components that have questionable quality)

Steven Gorsky

 


Posted by David Stambaugh (Member # 1102) on 06-25-2002, 11:38 AM:
 
Is the server dropping the client's connection to it? Or is the client losing its network connection independently? When this happens on the client, do you get a red "X" through the network icon on the lower-right of the task bar? This would indicate that the client thinks the network cable is unplugged or there's no link bit detected. Do the link bits remain lit on the switch at all times?

Are all the patch cables certified Cat-5? If they're home-made, make sure that pins 1 & 2, and 3 & 6 are each a color pair and consistently assigned at both ends.

Presumably, you've assigned private static IP addresses to each client and the server, making sure to use the correct subnet mask. Use a standard Class C address like 192.168.0.x with a subnet mask of 255.255.255.0 Make the server 192.168.0.2 and number the clients from 192.168.0.3 up to whatever the last one is.


 


Posted by Scott Norwood (Member # 30) on 06-25-2002, 11:41 AM:
 
Does this product even use IP? Or does it use some other protocol?

Have you run a packet sniffer on the network to look at the traffic? If you have a "managed" switch, have you checked its logs and statistics for clues?
 


Posted by David Stambaugh (Member # 1102) on 06-25-2002, 02:20 PM:
 
When the clients *are* working, is their network performance good? Any signs of communication problems with the server before the client gets disconnected, such as slow response times?

As Scott said, if it's a managed switch it would be helpful to check out the error statistics etc. on each port. Is the switch brand new, or is it used? If it's used it could be something weird with the way it was previously configured.


 


Posted by Mike Williams (Member # 1140) on 06-25-2002, 09:02 PM:
 
Can you ping any of the terminals when they "drop"?


 


Posted by Daryl C. W. O'Shea (Member # 1303) on 06-26-2002, 01:15 AM:
 
I've never seen a managed 16 port switch from LinkSys on any of my suppliers fax sheets, I believe they only make 24 port managed switches.

Since the clients are all running XP I'd still assume that it is an IP network since the protocols have been pretty much merged.

So far it's been my experience that most box office software is not protocol dependent and uses either mapped network drive (possible with most any protocol) or full UNC paths (also possible with most any protocol).

Cabling shouldn't be a problem given that clients are randomly dropping off the network yet being immediately reconnected on restart. This would suggest a software/hardware problem. If restarts didn't always work or work immediately I would then look into the cabling.

I guess we'll just have to wait for Mark to provide some more info.
 


Posted by Gordon McLeod (Member # 33) on 06-26-2002, 09:39 AM:
 
It is probably windows itself being flaky use reel operating systems
 
Posted by Mark Gulbrandsen (Member # 72) on 06-26-2002, 06:17 PM:
 
I have tried assigning static IP's to each computer, 169.254.104.1 (server), 169.254.104.10 through 15 for the terminals, with a subnet mask of 255.255.255.0. This does not work any more reliably than auto mode but logs on reel fast. I added IPX/SPX protocal and its family and things are much better. Auto assigning of IP's is a bad way to go as re-loggin in takes 2 min. or more. Re-log in in with static assigned IP's is real fast, about 15 seconds or less. Right now I have just about all the protocols and family that XP Pro offers and the system has been fine.....
The systems are based around the new Shuttle Micro PC's running Celeron 1.2ghz CPU's, 256 meg ram and 7200 rpm IBM drives. These things fly, and are only 8X8x9 inches! The switch is a Lynksys 16 port unit. No, the CAT 5 cables have not been integrity verified.
The software is Sensible Cinema Box Office For Windows... true 32 bit version. Any other suggestions are appreciated? We ahve several other installations running just fine. Same computers and all.

You old folks up in Canada still running on old 16 bit systems don't know what you're missing in the new 32 bit OS's, but then you are still using 1930's oiling techniques too so what more could one expect??
Mark @ GTS


 


Posted by David Stambaugh (Member # 1102) on 06-26-2002, 07:18 PM:
 
IPX/SPX is a protocol unique to Novell NetWare. I can't imagine how adding that protocol would improve things. In fact I doubt the protocol is actually being used at all even though it's installed (not saying adding it didn't help, only that I don't understand how or why). I do seem to recall though that IPX handles corrupted data packets better than IP. IPX's packet retransmit scheme is faster than IP I think. Way back many years ago we introduced IP to our Novell network and we immediately started having performance problems with clients that were running IP only. Found out that a 10/100 switch was randomly corrupting data on its uplink port. This bothered the performance of IP much more than IPX.

I think the finger is pointing at the patch cables, especially if they were terminated on-site, which is probably the case. Check those wire pairs very closely, and make sure pins 1 & 2, and 3 & 6 are color pairs and identically assigned at both ends of each patch cable (i.e. if pin 1 is solid orange at one end, make sure pin 1 is solid orange at the opposite end, etc.). I've seen supposedly knowledgeable low-voltage licensed technicians incorrectly terminate network cables by wiring them in progressive pairs straight through (1 & 2, 3 & 4, 5 & 6, 7 & 8 - WRONG). You want 1 & 2, 3 & 6, 4 & 5, 7 & 8 (although you can leave out 4 & 5 and 7 & 8).

I'm looking at a Black Box factory-made Cat-5-certified patch cable and the pin assignments are:

1 - Orange/White
2 - Orange
3 - Green/White
4 - Blue
5 - Blue/White
6 - Green
7 - Brown/White
8 - Brown


 


Posted by Scott Norwood (Member # 30) on 06-26-2002, 09:56 PM:
 
I'm with David on this one: buy/borrow/steal a cable tester and see what you find. Also try swapping thw switch for a known-good one.

And, Mark, some of us like our 64-bit OSes....
 


Posted by Mark Gulbrandsen (Member # 72) on 06-26-2002, 11:29 PM:
 
I have tested all the cables with a simple tester that confirms correct wiring of them. Although this is existing cable I suppose it could be a problem. Its is left over form some other POS system that was in use here. The Linksys switch is just a desktop type 16 port switch. I have an older 3-Com 10 port hub with me that I will try tommrrow. I donno if the hub can handle the 32 bit system though. Since we ahve sold other systems exactly like this that work fine I am beginning to suspect the switch may be bad. I also have a networking technician comming in tommrrow to check it all over. I'lllet you know whats found if anything.
I agree on the 64 bit stuff!! I wonder what they are thinking about all that 64 bit stuff North Of The Border.........?
BTW: There is no red X through the network symbol on the task , just a software warning that says to check the mapping to the files on the server, or to check the server.
Mark @ GTS


 


Posted by Steven Gorsky (Member # 603) on 06-27-2002, 02:07 AM:
 
Hubs (and switches) don't care if it's 32-bit or 16, they only are concerned about data speed (either 10 or 100Mbs). At this point I would say it is the cabling.

Ethernet networks require category 3 (or higher) UTP (unshielded twisted pair) for 10Mbs and category 5 (or higher) UTP for 100Mbs. If you're unsure about the rating of the wire that was already there, replace it, and the problems should disappear. If by chance it is cat 3 or 4 - you should be able to just change the network settings to use 10Mbs (not 100Mbs), and the problems should go away.

David did give one of the correct wiring standards (EIA/TIA 568B) for ethernet. I'm listing the other (EIA/TIA 568A) below for completeness.

1 - white w/ green
2 - green w/ white
3 - white w/ orange
4 - blue w/ white
5 - white w/ blue
6 - orange w/ white
7 - white w/ brown
8 - brown w/ white

Steven Gorsky


 


Posted by Jason Burroughs (Member # 68) on 06-27-2002, 03:13 AM:
 
How long are the runs to the switch?

Auto-assigning of IP Addresses is preferable, as it allows more flexability, and avoids IP conflicts. However there will be a delay on boot up, Unless you have a DHCP server on the network. What happens otherise is Windows listens for a DHCP server, times out, and then uses the Microsoft Private IP range 169.254.x.x

Annother thing to check is overall network traffic, see if there is a broadcast storm, you may have a computer that keeps forcing a bowser master election in there somewhere.

As far as multiple protocols, I would recommend either using IPX/SPX OR TCP/IP.


 


Posted by Steven Gorsky (Member # 603) on 06-27-2002, 05:43 AM:
 
After reading Jason's post above I wanted to mention a few things.

The maximum cable length for ethernet using UTP is 100 meters/328 feet. I only know of one theatre where this limit was exceeded, and fiber-optic cable had to be installed. While auto-assigning IP address is preferable on larger networks (50+ is about standard, I would say 20+), with a network this size, fixed IPs are the better solution. Broadcast storms shouldn't be a problem, again the network size used in theatres is pretty small, and a master browser election should be quick and only happen when a computer is brought onto the network (by powering up).

After Mark mentioned that the cabling was from the previous POS system, that made me 99.9% sure that's the problem - there is only a small chance that it is the hardware. It's either cat 3 or 4 and can't handle 100Mbs (short runs can unofficially), or it's something not even meant for ethernet.

Steven Gorsky


 


Posted by Gordon McLeod (Member # 33) on 06-27-2002, 06:54 AM:
 
well we have been using 32 bit when most of the rest were runnig 16 bit
 
Posted by Mark Gulbrandsen (Member # 72) on 06-27-2002, 04:35 PM:
 
The problem has been solved today. We moved the server upstairs today and ran all factory made cables over the floor to each station. There were still problems with the system set up that way and we traced it to the 16 port Linksys switch. The network guy also found that there were three bad CAT 5 cables in the system as well. All new cables are going to be pulled in tommrrow and certified. All terminals are running fine. So much for Linksys............
Thanks for everyones input on this situation!!
Mark @ GTS
 
Posted by Gordon McLeod (Member # 33) on 06-27-2002, 10:35 PM:
 
Funny never had a problem with Liksys up here
 
Posted by Mark Gulbrandsen (Member # 72) on 06-28-2002, 11:37 PM:
 
Actually the overall quality of the Linksys switch is VERY poor as comparing to a less expensive 3Com, or Netlink switch. The Linksys is in a cheap plastic case when the other two are in a nice metal case. Both the 3COM and the Netlink cost considerably less as well. A new Netlink Switch brought the system up and running in no time flat. I never had a problem with any Linksys stuff either till this episode, but I was not at all impressed with their declining quality and cheap plastic case....They also seem to have a poor reputation among ISP's around Aspen which is where I ultimately bought the Netlink switch.
Mark @ GTS


 






Powered by Infopop Corporation
UBB.classicTM 6.3.1.2