ds_seqgetnext: error in read() - Interrupted system call

[basiltarun]

Hello everyone,

I have an issue with the following error, reported in Datastage:

ds_seqgetnext: error in 'read()' - Interrupted system call

Here is the context:
- there are 16 input files to be processed, with identical layout.
- logic is simple, the Datastage jobs is suppose to extrat header/trailer and detail files and store them into separate sequential files
- the job uses 'grep ^' as filter command, to extract only the records that are needed for each output file type

The 16 input files are around 80MB (~ 400k records) each.

Tried the following approaches and still got the error above:
1. One sequence calling 16 instances of the same job, in the same time.
2. One sequence calling 16 instances of the same job, one at a time
3. 4 sequences calling 4 identical jobs (clones)
Any scenario above will cause a failure with the error mentioned in the top. A couple restarts on the failed sequence will have it finally completed successfully but what an inconvenience! In the 3rd scenarios, usually the first two sequences complete all right, but 3rd and 4th produces the error.

All this I can not reproduce in my development environment but happens only in production. I checked the input data, the thing is that every failure stops at a different record number!? Datastage v7.1

Any suggestions?

Thank you!

[ArndW]

Is the error in the sequential file read stage of the multiinstance job? What platform are you running on and have you compared your kernel settings on the development and production machines? NFILES or the like might be triggering the error.

[basiltarun]

Is the error in the sequential file read stage of the multiinstance job? What platform are you running on and have you compared your kernel settings on the development and production machines? NFILES or the like might be triggering the error.

Yes, the error is in the sequential file read stage. Sun OS is the platform. I can eventually have the settings checked, that's not in my hands...

However, even when eliminating multiple instances (scenario 3), the error still shows up. In that case, the same 4 (identical/clone) jobs are called for 4 times (within separate sequences).

[chulett]

You need to determine how the kernel and user settings differ between dev/test and prod. Or perhaps your uvconfig settings - something is different and the fix may be quite simple once you identify the culprit.

And 7.1? Get thee upgraded to 7.5.3!