InfoSphere DataStage PX job errors with no available output files

Symptoms

You receive the following error message:

Error Message: Sort_17,2: write failed: Output file full, and no more    output files [sort/merger.C:1393]
Message: Sort_17,2: Fatal Error: Tsort merger aborting:  mergeOneRecord() punted [sort/merger.C:1214]
Message: Sort_17,2: Fatal Error: Pipe read failed: short read [sort/m   erger.C:1740]

Causes

This error indicates that the job is running out of scratch, temporary, or swap space.

Diagnosing the problem

Checking the amount of space available without the job that is running can give a false indication that there is enough space.

Open the configuration file that is being used by the job. This can be found by checking the APT_CONFIG_FILE parameter of the job. You can find the configuration file in ../IBM/InformationServer/Server/Configurations/*.apt. Check the location of the scratch space that is identified for each node. While the job is running, issue the following command: df -k This command displays the amount of space that is allocated and being used on the system. Make sure that there is space available for /tmp, /swap, and the scratch area. Also make sure that the location that is identified in the configuration file for scratch space also has enough space.

Resolving the problem

Increase the disk space for /tmp, /swap, and /scratch. You can create an extra scratch area in your configuration file. For instance, you might add a second scratch disk area with the following command:

node "node1" 
{ 
fastname "gcp5bs03" 
pools "" 
resource disk "/opt/local/InformationServer/Server/Datasets" {pools ""} 
resource scratchdisk "/opt/local/InformationServer/Server/Scratch" {pools ""} 
resource scratchdisk "/etc/svc/volatile" {pools ""} 
}