Connection Drops after 3 mins

[Amit Anand]

Guys

I have an Application Imed Consent connected to my engine through TCP/IP.

The issue is whenever they have an HL7 to be delivered, it waits on the queue on their side and gets delivered to my engine after 2 hours sometimes.

The vendor says the he sees the connection drops after about 3 mins, although the thread stays always UP on my side. If I bounce my side at any time, the message gets delivered to my engine immediately.

I have verified, the configuration for this thread is exactly same as other 200 interfaces I have on my engine. I even deleted and rebuilt the whole interface, it did not help.

Any recomendations!

Thanks

Amit

Topic

[Rob Abbott]

Is this going through a firewall?  Firewalls like to tear down idle tcp connections…

Rob Abbott
Cloverleaf Emeritus

[Amit Anand]

Rob

No, both the servers are in the same domain.

Thanks

Amit

[John Mercogliano]

Amit,

  Based on past experience and other postings on this board it is probably the vendor’s problem.  Based on your description, after around 3 minutes the vendor attempts to release the connection but does not do it properly.  I bet if you do a netstat while the vendor side is queuing you will see your cloverleaf port in a fin_wait state.  This appears to be a common problem with vendors programs especially if they are developed on a windows platform.  

  Now, like a few systems I have, after they drop the connection as soon as something is queued they will attempt to reconnect but they can’t because you haven’t released the port yet.  Then after 2 hours which is the default tcp keep alive value the OS realized the the connection is not good any more and drops it.  The vendor is then able to reconnect.  

  So, hopefully it is just a vendor configuration issue and not a coding issue.  You do have an out if the vendor is unwilling to accept ownership and that is change the connection to multi server.  That will allow the vendor to connect multiple times while the other connections are waiting to time out.  But because there is a bug on there side you should limit the number of connections to like 10 or 20 to protect your system resources.

Here are a couple of links that have addition info.

http://clovertech.infor.com/viewtopic.php?t=3292&highlight=” class=”bbcode_url”>http://clovertech.infor.com/viewtopic.php?t=3292&highlight=

http://www.ncftp.com/ncftpd/doc/misc/ephemeral_ports.html” class=”bbcode_url”>http://www.ncftp.com/ncftpd/doc/misc/ephemeral_ports.html

http://clovertech.infor.com/viewtopic.php?t=2681&highlight=ephemeral” class=”bbcode_url”>http://clovertech.infor.com/viewtopic.php?t=2681&highlight=ephemeral

Hopefully this helps,

John Mercogliano
Sentara Healthcare
Hampton Roads, VA

[James Cobane]

Amit,

As a workaround, you could configure an alert to stop/re-start the connection if you have not received anything for x amount of time.  That way the engine can automatically cycle the thread to allow the vendor to re-connect.  But, I would continue to push with the vendor to look at their side to disconnect cleanly/fully.

Jim Cobane

Henry Ford Health

[Author]

Replies