Fatal Error: waitForWriteSignal(): Premature EOF on node: Socket operation on non-socket in IBM InfoSphere DataStage

Resolving The Problem

It is likely that the job log also has file errors related to file mapping.

Please ensure that temp and scratch file systems are NOT mounted with the "cio" option.

Internal DataStage code issues a mmap function to map into memory the lookup file that is created in your resource area. The resource areas are defined in the configuration files for each node.

We strongly recommend that all mounts that will be used for DataStage resource or scratch areas on ALL nodes be mounted with the options of "rw".


The cio mount option causes the file system to be mounted for concurrent readers and writers. I/O on files in this filesystem will behave as if they had been opened with O_CIO specified in the open() system call.  Using this option will prevent access in any manner other than CIO. It is impossible to use cached I/O on a filesystem mounted with the cio option. This means that mapping commands such as mmap() and shmat() will fail with EINVAL when used on any file in a filesystem mounted with the cio option. One side-effect of this is that it is impossible to run binaries out of a cio mounted filesystem, since the loader may use mmap().

Please review any DataStage configuration files and check the mount options on those filesystems used therein to insure they do not include "cio".  Once all the mounts on all the servers are changed and re-mounted, please try running the job again to verify the job.

Please ask your system admin to review the info above and check the mount options as described in the following:
mount Command