Message sent without waiting for ACK

[David Teh]

Hi folks,

On Cloverleaf 5.6 R2 on AIX 5.3.

In our test region, we had a source system sent a batch of 388 messages.

Inbound thread message count tallies.

Outbound thread “Outbound Message” log count tallies too.

All the messages were received and sent out in a 6-second period without any connection drop.

However, the Outbound thread “Inbound message” log count only has 124 messages. These are the ACK messages.

The GUI reflects similarly.

The Oubound thread is set to wait forever for an ACK. Rightly, I should have 264 messages pending in the queue.

What went wrong?

Topic

[Jim Kosloskey]

Do your outbound thread SMAT files (inbound and outbound) tally?

email: jim.kosloskey@jim-kosloskey.com 30+ years Cloverleaf, 60 years IT – old fart.

[David Teh]

Hi Jim,

For the OB thread, outbound SMAT file has 388 msgs. Inbound SMAT file has 124 messages.

Jim Kosloskey wrote:

Do your outbound thread SMAT files (inbound and outbound) tally?

[David Teh]

Hi guys,

Found the problem in an OB tcl.

Using a hex editor on the .msg shows 0d missing at end of each message.

[David Teh]

Hi folks,

Turned up the EO, and am pasting some parts of the log.

[“…” indicates portion I have removed]:

……

[pdl :PDL :DBUG/2:phxorder_to_phis:12/21/2012 13:03:23] PDL changed states: old 2, new 0

[pdl :PDL :DBUG/0:phxorder_to_phis:12/21/2012 13:03:23] Calling Tcl procedure: hci_pd.write

[pdl :PDL :DBUG/0:phxorder_to_phis:12/21/2012 13:03:23] with args: {{message message0}}

[pdl :PDL :DBUG/0:phxorder_to_phis:12/21/2012 13:03:23] Tcl procedure hci_pd.write returns ‘TRANSMIT’

[pdl :PDL :DBUG/0:phxorder_to_phis:12/21/2012 13:03:23] writing out 12433 bytes:

[pdl :PDL :DBUG/0:phxorder_to_phis:12/21/2012 13:03:23]

[Charlie Bursell]

[pd  :thrd:DBUG/0:phxorder_to_phis:12/21/2012 13:03:23] [0.0.25887353] Message successfully sent, not awaiting replies.  Removing msg from db.

You say you have Await Replies set.  I don’t see how.  Send this portion of the NetConfig for the thread in question:

{ REPLYCONTROL {

   { OBRESEND 1 }

   { REPLYGEN {

       { TYPE none }

   } }

   { TIMEOUT 45 }

   { WAIT 1 }

} }

[David Teh]

This is for that outbound thread:

   { REPLYCONTROL {

       { OBRESEND 0 }

       { REPLYGEN {

           { TYPE none }

       } }

       { TIMEOUT -1 }

       { WAIT 1 }

   } }

[Charlie Bursell]

You have something strange here,  It should wait forever for a reply

I would suggest you cntact the Support people.  Send them a copy of your NetConfig

[David Teh]

Ok thanks Charlie……

Charlie Bursell wrote:

You have something strange here,

[Bob Richardson]

One and All,

I am guessing here that the ReplyGen set to None disables the Await Replies as the engine *thinks* it doesn’t have to manage handling them.

Maybe experiment with selecting the Auto Resend and examine your EO log and see if AWAIT REPLIES appears.   The thread will never resend your message but then it may just activate the await reply logic.

Good luck!

[Jim Kosloskey]

Just a shot in the dark – but make sure your thread is set to Outbound Only.

email: jim.kosloskey@jim-kosloskey.com 30+ years Cloverleaf, 60 years IT – old fart.

[David Teh]

Ok thanks for the suggestions.

Will check further when I finally get back to office.

Hopefully support comes back with something useful.

[David Teh]

Still waiting for Support.

But just curious, what if the ‘0D’ is missing from the last segment of each outbound message (‘0D’ is present for the other segments)?

How would receiving systems generally react?

Any effect on Cloverleaf?

[Jim Kosloskey]

David,

You would need to find out from the rceiving system as to how they handle that.

It is entirely possible they just continue to collect data until they get something (like a full TCP/IP buffer) that would cause them to send and acknowledgement.

But since you know this is incorrect, correct the issue and retry. If your problem goes away, then it is a moot point as to how this particular receiving system reacts.

email: jim.kosloskey@jim-kosloskey.com 30+ years Cloverleaf, 60 years IT – old fart.

[David Teh]

Still waiting for support to revert.

But my further testings shows the following weird behaviour…

Without that particular TCL in place, the OB thread will wait forever for an ACK.

With other TCLs in place, it will also wait forever for an ACK.

With that TCL in place, all the messages were sent out even when no ACK was received.

Will update when support reverts.

[Author]

Replies