Skip to content

1.4.2 (Release Candidate 2)

Pre-release
Pre-release
Compare
Choose a tag to compare
@awsbuild awsbuild released this 20 Jan 11:38
1.4.2-rc2
098d0a1

Docker Images

coreoasis/model_worker:1.4.2-rc2
coreoasis/api_server:1.4.2-rc2

Components

oasislmf 1.4.7rc1
ktools 3.2.1

Changelog

  • Hotfix (rc2) - Remove previous output results on run error
  • Hotfix (rc2) - Fix Log and traceback storage
  • #222 - Update to model settings schema
  • #275 - Fixed delete operations with multipart as content type in swagger
  • #274 - Fixed reverting behaviour for complex models (custom gulcalc)
  • #281 - Added Task Queued state
  • #268 - Store run trace on success
  • #287 - Store ktools log directory in an Analyses
  • #283 - Fixed Worker not releasing memory from Python process

Feature Notes

Accessing Ktools logs

On an analyses run (pass or fail) if a ktools log directory is created the API will tar up <run-dir>/log and store as output under /v1/analyses/{id}/run_log_file/ (GET, DELETE)

Worker Job Status Update

Added two extra analyses.status states INPUTS_GENERATION_QUEUED and RUN_QUEUED

Analyses states overview:

  • NEW - Newly created analyses Object, portfolio successfully linked to a model.
  • INPUTS_GENERATION_QUEUED - Generate input files job sent to Celery queue.
  • INPUTS_GENERATION_STARTED - Generate input Job accepted by a Celery worker.
  • INPUTS_GENERATION_CANCELLED - Cancel signal sent, halting input files generation.
  • INPUTS_GENERATION_ERROR - Error while processing Input files, see /v1/analyses/{id}/input_generation_traceback_file/ for
  • READY - Input files generated, analyses ready for execution.
  • RUN_QUEUED - Analyses run sent Celery queue.
  • RUN_STARTED - Analyses run accepted by a Celery worker.
  • RUN_CANCELLED - Cancel signal sent, halting an Analysis's execution
  • RUN_ERROR - An error occurred during execution see /v1/analyses/{id}/run_traceback_file/ for details
  • RUN_COMPLETED - Run complete and results available at /v1/analyses/{id}/output_file/

Explicitly setting a custom gulcalc

From Oasislmf==1.4.7rc1 and onwards, custom models can explicitly set the drop-in component name.
The if the script/binary is not found on the current $PATH the MDK will raise an error.

CLI Example

$ oasislmf model run --model-custom-gulcalc MY_GULCALC

    ... 

COMPLETED: oasislmf.model_execution.bin.csv_to_bin in 0.02s
STARTED: oasislmf.model_execution.bin.csv_to_bin
COMPLETED: oasislmf.model_execution.bin.csv_to_bin in 0.03s
STARTED: oasislmf.model_execution.bin.prepare_run_inputs
COMPLETED: oasislmf.model_execution.bin.prepare_run_inputs in 0.0s
STARTED: oasislmf.model_execution.runner.run       
Run error: Custom Gulcalc command "MY_GULCALC" explicitly set but not found in path.

For worker deployment add the following to an oasislmf.json file

{
    ... 
    "model_custom_gulcalc": "<MY_GULCALC>"
    ... 
}
Assets 3
Loading