Unix error=111,APT_Communicator::connectTo: connect() fail

wbjk · Post by **wbjk** » Wed Sep 12, 2007 9:16 pm

//In the enviroment of my company,,there are 40,000 jobs running on 20 servers per day. I think maybe datastage run out of the tcp resourcs of the system? how to fix it?
//CCB_Decode is a cutom stage.
//
here are the details of log:

2007-09-13 02:28:46 FATAL event id: 17535
CCB_Decode,3: APT_Communicator::connectTo: connect() failed due to Unix error = 111 (Connectioo
n refused) on node ODSETL13 using ConnectionInfo object 'TCP, connection Host: localhost.localdomaii
n (127.0.0.1), TCP port number: 11000', RETRYING connect()
APT_Communicator::connectTo: connect() failed due to Unix error = 111 (Connection refused) on
node ODSETL13 using ConnectionInfo object 'TCP, connection Host: localhost.localdomain (127.0.0.1),,
TCP port number: 11000', RETRYING connect()

2007-09-13 02:28:46 FATAL event id: 17536
CCB_Decode,3: Fatal Error: APT_Communicator::pmSendPartitionInfo() failed on node ODSETL13 forr
partition 3 of dataset 0 (write failed to handle 4) Bad file descriptor

2007-09-13 02:28:46 FATAL event id: 17537
node_node0005: Player 1 terminated unexpectedly.

2007-09-13 02:28:46 FATAL event id: 17538
main_program: Unexpected exit status 1
Unexpected exit status 1

2007-09-13 02:28:51 FATAL event id: 17539
main_program: The Section Leader on node node0004 has terminated unexpectedly.

2007-09-13 02:28:51 FATAL event id: 17541
main_program: Step execution finished with status = FAILED.
2007-09-13 02:28:51 STARTED event id: 17543

ray.wurlod · Post by **ray.wurlod** » Wed Sep 12, 2007 11:33 pm

The APT_Communicator class is used mainly (only?) for communication between player processes, which basically means when repartitioning is going on. Perhaps you could try to reduce the amount of repartitioning that is occurring in your job design.

wbjk · Post by **wbjk** » Thu Sep 13, 2007 12:34 am

I am not a premium poster, I cannot see all the content of your post ,can you give me a private message?

This error has happened for some days. So many core files appear in the user's directory!

ray.wurlod · Post by **ray.wurlod** » Thu Sep 13, 2007 1:27 am

For less than 30c per day you can be a premium member, and help to defray the bandwidth costs needed to keep this site alive. I do not propose to prejudice that mechanism.