Closed nsheff closed 1 week ago
It appears as though it is ignoring the pipeline_name given in the interface and using the one in the pipestat output schema.
yes, adding output_schema: output_schema.yaml
to my pipeline_interface was what it was looking for.
I believe this is now solved with the recent releases of Looper 1.8.1 and Pipestat 0.9.3. With the refactoring, Looper prioritizes the pipeline_name in the pipeline interface which can be passed to pipestat via the generated pipestat config file or as a parameter to pipestat. Both of these methods now take priority over the pipeline_name found in the output schema. Looper will now warn the user if there is a pipeline_name mismatch and default to the one provided in the pipeline interface. The docs have also been recently updated as part of those recent releases.
How do you configure the pipestat pipeline name when passing it via looper?
Here's my looper config;
The problem is, before any pipelines are even run, looper is "helpfully" creating the
results/pipeline_stats.yaml
file... with this content:Then, when the first job runs, it tries to report to that file with a correct namespace,
add_to_seqcol
, and that gives an error:Because 'default_pipeline_name' is already being recorded in that file, and then, understandably, when I try to report a different namespace, it's unhappy. So, the configuration is bad.
I tried adding
pipestat.pipeline_name: add_to_seqcol
but it does not pick it up.