- All KBase Jobs use a CG to schedule jobs
- The default CG is
njs
- Changing a CG will affect the scheduling of all future jobs
- If you don't know what you are doing, please ask for help in #devops , #sysadmin, and/or #poc
- The catalog CG UI is available at https://narrative.kbase.us/#catalog/admin
- The default resources at kept in ee2
As of Dec 2022, the default resources are
{"client_group": "njs", "request_cpus": "12", "request_memory" : "10000"}
{"client_group": "bigmem", "request_cpus": "16", "request_memory" : "250000"}
{"client_group": "extreme", "request_cpus": "32", "request_memory" : "200000"}
- To get the
Module Name
is easy, its just the first line in your kbase sdk app spec egmodule kb_StrainFinder {
- To get the
Function name
is harder, you have to go into the App/UI definition that you want and see what method it calls
- To get the default settings and run on different hardware, set the CG as follows:
{"client_group": "bigmem"}
{"client_group": "extreme"}
WARNING You must know the actual capacity of the hardware or the job will be stuck forever and never run
{"request_cpus": "64", "client_group": "bigmem", "request_memory" : "500000"}
WARNING THE DTN NODE IS FOR UPLOAD JOBS, NOT FOR JOBS THAT DO SIGNIFICANT COMPUTATION WARNING WARNING PLEASE CHECK WITH THE DEVELOPER BEFORE ADDING COMPUTATIONALLY INTENSIVE JOBS TO THE UPLOAD QUEUE WARNING
- Get module name and function name
{"client_group": "kb_upload"}
Heres an example
kb_model_analysis model_heatmap_analysis kb_upload /dtn/disk0/bulk/${username} /staging (read-only)