Error 3043 - Network Connection Interrupted

[irish634]

Certain times and various users, I have front-end users lose connection to the back-end. They end up with error #3043 that says: "Your network connection was interrupted. To continue, close the database and open it again." -If this is a new record when this occurs, all data entry is lost and of course my auto numbering gets screwed up.
Even having a persistent connection does not alleviate this.

I have been able to prolong the time it takes for this error to happen by changing some timeout settings, but I am curious about something.

Is it a good idea to have an unbound form and then when the user is ready to submit the data they have entered, dynamically refresh the links to the back-end?

 

[spikepl]

I reckon you could try your refresh thingamajig in the before update event of a bound form - any other opinions ?

However to "my auto numbering gets screwed up" my reply would be: if that is a problem then your design is not the way it should be. The autonumber in Access is essentially not for human consumption, and so who cares what the value is. If you want sequence then roll your own.

 

[Uncle Gizmo]

Has someone stuck a wireless link in your network somewhere?

 

[vbaInet]

Has someone stuck a wireless link in your network somewhere?

Exactly what I was going to ask.
They must be accessing this over an unreliable network, such as a wifi or there's a problem with the NIC.

 

[spikepl]

I have an installation on a 1 Gbit cable that does this once or twice a day so ...

 

[vbaInet]

I have an installation on a 1 Gbit cable that does this once or twice a day so ...

Speak to your Network Engineers

 

[irish634]

However to "my auto numbering gets screwed up" my reply would be: if that is a problem then your design is not the way it should be. The autonumber in Access is essentially not for human consumption, and so who cares what the value is. If you want sequence then roll your own.

Agreed....
I guess what was trying to show getting something other than desired.

Has someone stuck a wireless link in your network somewhere?

No.

a problem with the NIC.

Yes, it is related....
http://support.microsoft.com/kb/2740020
I've also made some registry changes to prolong the timeout to the duration of the shift.

 

[spikepl]

Speaking to meself is utterly boring and, further, useless

 

[vbaInet]

Speaking to meself is utterly boring and, further, useless

Ha, sometimes it's healthy.

@irish634, have you been able to spot a pattern of when it happens? I.e. is it random or after unlocking the computer or when the application gets focus?

To answer your question about using an unbound form, that's up to you but that means you would have to write much much more code to handle pretty much everything. With your current setup you could always test the connection before attempting to commit the data, i.e. Before Update, Before Insert, Before Delete, and if needs be attempt to relink the db then take the necessary action.

 

[irish634]

@irish634, have you been able to spot a pattern of when it happens? I.e. is it random or after unlocking the computer or when the application gets focus?

To answer your question about using an unbound form, that's up to you but that means you would have to write much much more code to handle pretty much everything. With your current setup you could always test the connection before attempting to commit the data, i.e. Before Update, Before Insert, Before Delete, and if needs be attempt to relink the db then take the necessary action.

Before I modified the registry and followed the instructions in the KB article I posted, the PC would not even go to sleep. If the front end was open and the user filling out the form, the connection could be lost in as little as 4 minutes.

As far as the before update, I already have a procedure to check links on startup. I could conceivably modify that to auto-relink if needed. I was just wondering how others might or might not handle this situation.

 

[vbaInet]

You might be in a bit of an unfortunate situation. Do you think that there's some code/application somewhere, perhaps on the network, that's re-creating/refreshing the UNC path to your backend? Otherwise it may just be an unfortunate problem with the LAN. Maybe consult with a Network Engineer for suggestions.

 

[irish634]

You might be in a bit of an unfortunate situation. Do you think that there's some code/application somewhere, perhaps on the network, that's re-creating/refreshing the UNC path to your backend? Otherwise it may just be an unfortunate problem with the LAN. Maybe consult with a Network Engineer for suggestions.

I don't think so. The only code for linking currently in any of my front-ends is initiated by manually selecting the back-end at startup when the connection check fails. I can certainly write code to check for the connection before_update.

As far as applications and connections go, Access is the only app that is affected. If I open a spreadsheet, ppt, word doc, or text file in the same network folder and from the client workstation, I don't have the issue. Like I said, I've been able to prolong the connection error to anywhere from 4-6 hours by modfying the registry and following the kb article but the issue still comes up.

I suppose I could also write something to check the connection in the On_Timer event if I need to. Just looking for others thoughts.