A pipeline is a seriez of applications running one after the other. Any
two connected applications are related either by files or database.
Currently we support only file-based connection. The pipeline jobs may
fail if any step inside the pipeline fails, or the files that should
connect two applications are not generated correctly, due to pipeline
misconfiguration.
For example, in the phylip tree protocol, where six steps (Clustalw
-> SeqBoot -> ProtDist -> Neighbor -> Consense ->
DrawTree) are generally involved, the failure of "seqboot" will stop
the entire pipeline. Or, "consense" should generate a phylip-tree
format file for drawtree, but in fact the expected file is not found
because pipeline designer chose the wrong configuration for "consense".
In this case, there is no failed job inside pipeline, but the pipeline
itself hangs and there will be no further jobs to be initialized.
Continue reading →