Engine not recognizing lost connection

[Rich Durkee]

We have this one thread that receives data from a process on an old Windows NT server. Every once in a while the Windows server process will stop or get hung up. The Engine does not recognize this problem. The conection is Up and green, but no messages are received. I know I could set up an alert to monitor for messages not received in a certain amount of time, but that doesn’t always work well for inbound threads that send groups of messages a irregular times with no messages in between. Naturally the Windows Server user should notice that there are messages backing up on their server but they don’t. They think it is the Engine’s responsibility.

Anyone else see this “problem” and what was your solution?

Topic

[Bill Bertera]

Does the Windows server realize the connection is lost, and then try to resend, but can’t because cloverleaf still thinks the old connection is up? Basically, does recycling your thread fix the problem?

If so, you can configure your TCP listener to multi-server, that will allow multiple TCP connections. Eventually the lost connection will be dropped, assuming your Cloverleaf OS has TCP_KEEPALIVE running.

[Charlie Bursell]

Are you doing any Tcl “exec” calls?

There is a known problem, especially with Windows 2003, with STDOUT in Windows 2003 and Tcl 8.3.  It has to do with the way Windows buffers STDOUT

A work-around is to make sure you always do something with the STDOUT or STDERR return of an exec call.

If you do not need the returned data, do something like:

exec ……..  > nul:

If you do need the returned data

exec …… > tmpFile

set data [read_file -nonewline tmpfile]

file delete tmpfile

There is a lot of chatter going back to the late 1990’s about this on comp.lang.tcl.

When Tcl 8.3 was released, Windows 2003 was not even on the drawing board.

[Dennis Dunn]

Hello,

I had this same problem, my HL7 message source would connect, send messages and then disconnect but the thread would stay in the UP state. I expected it to go from OPENING to UP and then back to OPENING.  

I discovered that I had two .cli files in the %hicsite%/exec/processes/ directory.  I had renamed the process between starts and this left an extra .cli file in the process direcotory.  

I deleted the extra .cli file and network monitor is showing me the expected behaviour.

HTH,

–dennis

[Dennis Dunn]

Hi,

I’m afraid that I jumped the gun.  I did some more testing and the information in my previous post appears to be bogus.  Sorry.

I wonder how long it will take before the connection closes?

–dennis

[Chak Kwok Yin]

Hi

I would like very much to test out William Betera advise at some of the sites inbound threads. I tried to configure an inbound thread (server) to multi-server but received this error message upon startup.

04/04/2007 16:30:19 [tcp :init:ERR /0:   NSAP3_ADTP] Multi Server is not licensed, can not start thread.

Anyone face this restriction even though our company paid for each connections and its maintenance?

Regards

Chak

[Jim Kosloskey]

Chak,

I think there is no charge for the license, but you need to contact Quovadx to get a separate license key for that functionality.

Jim Kosloskey

email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.

[Russ Ross]

Jim your recollection is the same as mine because I had to get that extra license key which is free but only if you ask for it and it is called

cl-aom-server-socket

Russ Ross
RussRoss318@gmail.com

[James Mestack]

noticed post from Bill Bertera:  How do I tell or configure my Win 2K3 server to have TCP_KEEPALIVE running?

[Bill Bertera]

James Mestack wrote:

noticed post from Bill Bertera:  How do I tell or configure my Win 2K3 server to have TCP_KEEPALIVE running?

I think its a registry setting. Search for “KEEPALIVE” or “TCP”.

[Michael Hertel]

Here’s a previous post that might work:

https://usspvlclovertch2.infor.com/viewtopic.php?t=734&highlight=keepalive

[Author]

Replies