Issue

Server jobs operate in a largely sequential manner, and can be designed to ensure that data reaches particular stages in a job in a desired order. This allows designs where a single Sequential File stage may contain multiple input links writing to the same file, and this will behave correctly, predictably and repeatably.

Due to the way Parallel jobs operate, data flows through the job as it becomes available, resulting in a target file where the ordering of rows is not consistent or predictable.

Actions

Parallel jobs that replace Server jobs involving concurrent write operations should be resigned to collect and order the results in a deterministic manner before writing to the Sequential File using a single input Link.

See also

• Sequential File stage