Process hangs while sending some large batches

[Krishna Kumar]

Some time we are experiencing a problem sending some large batches over our charge Interface. It seems inbound thread finishes reading the batch from specific folder but the out bound thread hangs turning red showing pending messages. I have to stop and start the out bound thread to initaite the flow of messages again.

Here is the set up for outbound tab in the outbound thread:

Reties : 3

Interval : 12

Edi Batch : None

Await Replies: Selected

Time out : 0

TPS Inbound Reply : check_ack TPS

Any body has any suggetion to fix this problem?

Thanks

Topic

[Jim Kosloskey]

Check the thread very carefully to see if messages are not really flowing. You can do that by having the status in the NetMonitor GUI for the thread refreshh (Update).

Check the outbound message count and see if it is not incrementing. The pending count may not go down but increase even if messages are going out.

This is most likely happening because the arrival rate from the file is higher than the receiving system can handle the messages.

You have some control in that you can cause the fileset (inbound) thread to pace how often it reads and how may it reads. That may balance the equation out for you.

There is no hard and fast rule for these settings as it is really dependent with how efficient your integration is and how much load the receiving system can handle.

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

[Krishna Kumar]

Here is the situation:

Outbound message count increases once time comes up in the inbound thraed, message count will remain same for hours. I do not think that is normal.

I have set a set up in my inbound thread schedule like

Read intaerval : 5

Max Messages   : 100

These messages are going from Cloverleaf to EPIC. I can see  EPIC monitor waiting for messages to be sent. There also I can see wait time increases as the process hangs.

Do I have to set up sendOK to destination since it may be waiting for this from me before accepting next message making time out to 10 0r 20? Right now time out is set to 0 that is infinity.

Once cloberleaf help desk suggested setting up recovery 33.

I do not know which one will fix this issue

[Jim Kosloskey]

Krishna,

I assumed you had reply handling (recover_xx or auto reply if on 5.6 or beyond) in effect.

It is very possible that you are waiting on a reply. Since you have wait forever set that is what it will do.

Sometimes receiving systems cannot keep up and then don’t reply – or the reply is ‘lost’ .

If you check your SMAT (inbound and outbound for the outbound thread) you can tell if you received a reply for the last message sent.

If you set a timeout value, then the first pending message will be resent and that can also go on forever if the receiving system is not replying. In that case the outbound count will go up but the pending count will never reduce.

You need to determine if you are really waiting on a reply (I suspect that is the case) and why there is no reply forthcoming from the receiving system.

Once you know that you can proceed to determine how to set up the handling.

I always have reply handling defined in my realtime interfaces (recover_xx) and I suspect you may need that as well.

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

[Sundeep Kumar]

If you do a hcidbdump -r -d you should see one message in state 14. You will need to delete the message, save it and let the messages flow. Howver the correct sequence for doing this is:

1. Stop the remote server that is getting the data.

2. Stop the out thread ie the client thread sending the date from Cloverleaf to Epic. ie the OUT thread

3. Then hcidbdump -r -s 14 -d and save the message and delete it from queue

4. Start the remote server port

6. then start your OUT thread.

Chances are your remote server errors on something, recovers but never sends back an ACK so your OUT thread is left tryibg to resend it over and over again

[Author]

Replies