Datasocket bug in Win2K -- stops updating

[Doug R]

It seems datasocket has problems with large data packets. The more data you send the sooner it stops.

[Morag C.]

YES!! I have this problem too and no-one seems to have an answer to it. I have just filed a comment under another part of Knowledge base but I will reproduce it here. I have seen the problem under Windows 95 and Windows 2000; both on the same timescale (loss of reader connection after ~7 hours). I have approached NI (UK) several times and they can reporduce the problem but have not managed to fix it (tried many different builds of the DSS and am now running LV 6.0.3, tried connecting in autoupdate mode, 'manual' update mode and so on and still have the problem). Anyway, here's a description of it.

I am running the datasocket server, writer VI and reader VIs on a switched subnet in Germany the reader VIs in Scotland. There is no data loss between reader and writ
er if both VIs are run in Germany but the reader loses connection to the datasocket server after about 7 hours continuous connection when I try to run it in Scotland. It displays the error code 30301. I have tried doing as Jeff M. suggested (didn't work) and also tried a knowledge base suggestion to temporarily reconnect to the datasocket server with a blank URL (reconnect.vi). This also didn't work. The only solution I have found is to stop and restart the entire VI. I think that there is a clue in the fact that I do not see this problem when trying to run the reader and writer within the same subnet but only when I separate them on to different subnets. It's as if a buffer somewhere is overflowing after 7 hours and stopping the connection. By restarting the VI, the buffer is cleared and data continues to arrive. It might be that the number of TCP/IP packet collisions is much higher when the reader and writer VIs are on different subnets and this is why the 'buffer' overflows.
Any s
uggestions welcome!

[Doug R]

I took your programs and put them on various computers in my factory. After tests
Server client status
4.0 XP Fail ( ~ 2 hours)
2k XP Fail ( 4 hours )
XP XP Fail ( 5minutes )
XP 2K Fail ( 35 minutes)
XP NT running
2k NT running
2K 2k not tested with your program
shown bad with mine

etc etc etc

I think I can safely say it is not limited to any one OS... although I can't try win98 since my disks are NTFS.

Tried XP seriously yesterday... major ordeal and ended up seeing the system would not run for more than 20 minutes before bombing ( this is with my DSC system ) Today I am back to 2K on the DSC system running > 6 hours wi
th no problems.

If the server stops giving data and four clients stop getting the data.. you can reset any ONE client to get them all to run.

[Doug R]

NI HELP

datasockets is Broken. Does not work and may have never worked for large scale applications. LOGOS is just as bad.

I was talked into DSC since it didn't have these problems ( using web server ) but this method has some drawbacks.

[josborne · Accepted solution]

NI has confirmed that the problem that I originally posted here (Datasocket bug in Win2K -- stops updating) is a BUG in Datasockets. The problem is that COMPILED code that uses the OLD-style datasocket VIs which were used before LV6 (the ones with connect/disconnection and are availbe in vi.lib\platform) use the "wrong version" of underlying code, or something like that. They provided me with a separate fixed DSTP installer that fixed this problem.

For a work-around without the fixed version of datsockets, they reccomened to first use one of the new LV6-style datasocket VIs in your code to read a value from the URL before connection to it using the old-style VIs. This will force the datocket VIs to use the "correct" code. I haven't tried
this technique since I have the fixed datasockets installer.

P.S. I don't think NI has any knowledgebase article about this bug yet.

[Ian R]

We a have system with 4 PCs writing data to a server PC via datasockets using LabVIEW 6.1 and the new datasocket VIs and are seeing a similar problem. The server stops receiving data after running anything between 2 days and 2 weeks.

The server is running Windows 2000 server edition and the other PCs Windows 2000 Professional, all at service pack 2 (they are connected by a Gigabyte network link). The server is a dual processor machine and runs a LabVIEW application and Oracle.

Each PC is writing around 400 floating point values to the server each second.

Is this two much data for datasockets or is there a bug in datasockets or Windows 2000.