Hi
I have a controller WB ans 5 runners running on fme server. Controller passes params to runnrs unis FMEServerJobSubmitter.
Looing at the servers completed jobs I see 1.4 MILLION finished jobs - all of them runners. This is in the last 24 hours - the very last page is still in the last 24 hours.
See the image for info. 1st page
last page
Only explainaition I hgave is the server creates a job for every feature processed, rather than for every WB.
Why is that?
Yes, that's true. If your controller workspace reads 1.4 million features, and passes each to an FMEServerJobSubmitter, then that job is going to be submitted 1.4 million times.
If the workspace name is held as an attribute on the incoming features, and you only wish to run each workspace once, then use a DuplicateFilter transformer in the control workspace to filter out excess features with the same workspace name.
I guess really the question is what are these 1.4 million features and why are you reading them in the control workspace. The control workspace should be for starting the transformation, and shouldn't need to be reading the actual source data.
Does that help?
Yes, this looks like a FMEServerJobsubmitter called for every feature in a master workspace. Transformers are executed per feature, so if you want the FMEServerJobsubmitter only executed once, you need to use a Sampler to sample only the first feature to execute the FMEServerJobsubmitter.
On the side: It is pretty cool to be able to run 1.4 Million workspaces in 24 hours. Great performance!
I havent done the controler and the worker. I am trying to understand someone alses work. Not very easy.
What puzzels me is that the controller should only pass limited amount of features to the worker.
The controler reads a table (not spatial) and gets some IDs, - 1897 of them. It then selects some geometries from a spatial table using feature reader and matcing on IDs. Then these geometries are passed to the runner to be intersected with other geometries from another spatial table (all oravle spatial).
I was expecting a low number of jobs in this case, I am not sure where 1.4 mil features are coming from.
I havent done the controler and the worker. I am trying to understand someone alses work. Not very easy.
What puzzels me is that the controller should only pass limited amount of features to the worker.
The controler reads a table (not spatial) and gets some IDs, - 1897 of them. It then selects some geometries from a spatial table using feature reader and matcing on IDs. Then these geometries are passed to the runner to be intersected with other geometries from another spatial table (all oravle spatial).
I was expecting a low number of jobs in this case, I am not sure where 1.4 mil features are coming from.
I think we'd need to see your workspace (a screenshot at the very least) to help identify what is happening. Why not disable the FMEServerJobSubmitter (select it and press Ctrl+E), turn on feature counts if they aren't already, and run the workspace again? That'll tell you (and us if you post a log/screenshot) exactly what is being read. My guess is that the FeatureReader is returning more features than you expect. Maybe some of the filtering (where clause, spatial filter) is no longer working, either an FME issue or perhaps the source data changed?
Before I go into posting parts of the WB, is there a way in fme server to filster the job log and find the controller entry?