LAE

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: lae_errors

This post has been created for discussion of error code brain.node.http.transportProtocolError.

Please reply to this thread with any description of error conditions, diagnostic information, and recommended resolutions.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rgvenkatesh

Hi,

I am hitting a Http node using a POST call with 7K input requests, the node fails with the following error and stops the flow with a cross mark on the top of the node:
"A connection error occurred for POST request to http://xxxxxxxxxxxxxxxxx: Details: java.net.SocketException: Connection reset."

I understand this might be connection issue but what I am worried of is, at the time of the error the node has already committed close to all the requests in the backend database and when I re-launch the node again with the same input, all of them fail with uniqueness error.

I am hoping there should be a way to handle the http node transactions wisely and have an advantage of tracking such connection issue on requests basis, therefore if the connection issue comes on the 6999th request I should be able to track the connection error in the output response and continue with the next request instead of the node failing altogether.

Please help.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: stonysmith

To help diagnose the situation, right click on the HTTP node and go to the Status tab.
Copy the name of the Output temp file name.

Paste that filename into a BRDFile node, and on the Optional Tab, select "NoWrite=true"
Run that input node and see what the record count is.. it should tell you how many records were successful, which will be one less than the record which failed.

Also, you can then run a filter on the Response Code field to find out if there were any other errors.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rgvenkatesh

Originally posted by: stonysmith
					

To help diagnose the situation, right click on the HTTP node and go to the Status tab.

Copy the name of the Output temp file name.


Paste that filename into a BRDFile node, and on the Optional Tab, select "NoWrite=true"

Run that input node and see what the record count is.. it should tell you how many records were successful, which will be one less than the record which failed.


Also, you can then run a filter on the Response Code field to find out if there were any other errors.
			

Hi,

I see we can track the number of requests which went through fine with the BRDnode and temp output file, but my question was more on the httpNode wherein
the node stops with the red cross mark because of the connection reset issue and what I was expecting is,
if we re-run the node after the connection breaks it should pick up from where it left before failing.

Not sure If am putting up the queries correctly but I hope you understand the scenario here.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: stonysmith

I understood your request, and we're researching the proper answer.
I was just suggesting that you look at the output temp file as a method to determine which record failed.

Another question.. does it work properly if you break the 7000 records up into some chunks?
I am wondering if the remote server you are connecting to has a max time for a session to remain open.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rgvenkatesh

Originally posted by: stonysmith
					

I understood your request, and we're researching the proper answer.

I was just suggesting that you look at the output temp file as a method to determine which record failed.


Another question.. does it work properly if you break the 7000 records up into some chunks?

I am wondering if the remote server you are connecting to has a max time for a session to remain open.
			

Hi,

I have a max of 10 threads configured in my graph which cannot be increased.
Also, the total number of requests coming in are 70K which are divided by a hash split (10 outputs) , so one HttpNode has 7K as input.

And yes, even I was thinking about the remote server time out issues because on real time we would be hitting the httpNode with a max of 7K requests but even then I wanted the node to not stop but to emit out the connection reset on the output XML so that can be tracked as a error on DB and be re-launched at a later time.

So before contacting the remote server team, I just wanted to explore on Lavastorm to see if we have a way to go around this and see if the httpNode can understand the connection time out errors as a logical error and not as a system error which leads to control flow being stopped.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rgvenkatesh

Hi,

I see that the Http Node does not give out the outputs real time whereas it waits for all the input requests to be processed and then emits out all the responses at once.Is this a limitation in Lavastorm?

Question:
We are passing 7K requests into a HttpNode and in the current situation the node processes all 7K at once.
How do we process one request at a time, update the output on the a DB table and iterate back to the next request (similar to functional call in while loop)?

Is there any other to achieve the above without using a HttpNode?

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: stonysmith

Originally posted by: rgvenkatesh
					

I see that the Http Node does not give out the outputs real time whereas it waits for all the input requests to be processed and then emits out all the responses at once.Is this a limitation in Lavastorm? 
			

There's something odd there., I've always seen the HTTP node process each record individually.

What version of LAE are you running and what version of the LAL library do you have?
You can click on View/Dependent Libraries and paste the full text here.

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: rgvenkatesh

Originally posted by: stonysmith
					

There's something odd there., I've always seen the HTTP node process each record individually.


What version of LAE are you running and what version of the LAL library do you have?

You can click on View/Dependent Libraries and paste the full text here.
			

Hi,

I believe the HttpNode is processing the requests one by one internally, but what I mean is, when it emits out the output , instead of sending the outputs one at a time it waits for all the requests to be processed and then sends out the response all at once.

There are the dependent libraries used:
C:\Program Files\Lavastorm\LAE6.1\lib\brain\brg\core.brg - LAE 6.1.0 #14
C:\Program Files\Lavastorm\LAE6.1\lib\brain\brg\inFlow.brg - LAE 5.1.0 #1
C:\Program Files\Lavastorm\LAE6.1\lib\brain\brg\lal1.brg - 6.1.48.0.19
C:\Lavastorm\Lavastorm_613\Libraries\utils.brg - 1.1
C:\Lavastorm\Lavastorm_613\Libraries\soap_xml2.brg - 1.3
C:\Lavastorm\Lavastorm_613\Libraries\Text_File.brg

Example:
7K request -->HttpNode (Processing one record internally) -->7K responses all coming at once instead of one at a time.