Changelog
3.0.0
New features
Use ScienceDataModel for artefact management by default.
Update run and run-from-sdp subcommands to accept SDM
Initialise SDM stage to support run subcommand.
Update target calibration workflow to also include rapthor-di calibration.
2.1.0
Use default values for configuration having values defined by strategy for all INST pipelines.
Update default values for some continuum imaging parameters.
2.0.2
Fix piper dependency
2.0.1
SKB-1312: Disable delay calibration stage in instrumental calibration pipeline
2.0.0
New features
Added new steps inside "target_calibration" stage:
complex_gain_calibration
ionospheric_calibration
Use Rapthor's new "slurm_static" mode when running e2e on multiple nodes.
Use Rapthor for DD calibration as DI calibration will be done using complex_gain_calibration.
Add Multi-scans input support for single source of calibrator and target observations.
Support additional configuration for CIMG stage.
Breaking
"max_selfcal_loops" parameters is now part of "self_calibration" nested configuration.
Update Json config contract for
run_from_sdpsub-command to match with pipeline stages.Update config contract for
runsub-command to pass multiple ms file paths for given calibrator source.Additional configuration for CIMG stage.
General
Update all stages as per the ska-sdp-exec-piper V2 API
Disable resource monitoring in all of the sub-pipeline runs.
Added support for INST v0.8+.
1.0.0
SKA SDP E2E CLI
Create a python-based CLI for running batch end-to-end pipelines. More details on the CLI are in RTD
Breaking
Batchlet and batchlet plugins modules to new repository at https://gitlab.com/ska-telescope/sdp/ska-sdp-exec-batchlet.git.
Remove streamlit sdp interface module.
0.4.2
Batchlet
Add a new option
dask_report_dirwhich enables and stores dask performance reports. More details are in the batchlet documentation
0.4.1
Batchlet Plugins
Read hostname from logs, and include it in the benchmon csv
Batchlet
Remove monitoring results for failed pipeline runs.
0.4.0
Batchlet Plugins
Added batchlet filter plugin for SKA SDP standard logging format
Added consumer plugins for SKA SDP Configuration database (etcd) and simple csv file
The plugins are part of the
ska_sdp_batchlet_pluginspackage. Detailed documentation is available here
Batchlet
Add support for log monitoring in batchlet
Add support for adding custom filter and consumer plugins, used for monitoring of the logs
(Breaking change) Update the
monitorsection of the input configuration schema. Please see the configuration details in the batchlet documentationThe
DaskSlurmClusterclass can take a decision appropriate parameters for the dask cluster based on allocated/available resources, similar toLocalCluster. Detailed documentation is available here
E2E Scripts
Added Rapthor as part of e2e scripts
The
e2e.shnow calls other individual pipeline scripts as subprocesses.Added suppport for completely spack based execution, so user does not have to clone the scripts from git. The steps to run the scripts are updated in the documentation
0.3.2
Batchlet
DaskSlurmCluster now inherits
dask.distributed.SpecClusterclass. This allows DaskSlurmCluster to be used in manner similar to other clusters in dask ecosystem (LocalClusterofSSHCluster). This also adds ability to gracefully close the cluster upon exit.Add support for
ska-sdp-benchmark-monitorin non-slurm (local) execution ofbatchlet.
E2E Scripts
Write output of instrumental calibration pipeline in the same directory as
OUTPUT_DIR, instead of creating a unique timestamped subdirectory.
0.3.1
Batchlet
Force dask slurm cluster to start scheduler on random port, and use a unique scheduler file
Allow batchlet to start a local cluster when slurm environment is not available
Fix Typer's click dependency issue.
E2E scripts
Update slurm scripts to use spacks with metamodules
0.3.0
Add
batchlettool to run batch pipelines. The documentation for batchlet is available on readthedocs.Removed
batch-continuum-imaging-pipelinecli tool, it is now replaced bybatchlet.The
ska_sdp_e2e_batch_continuum_imagingpackage now contains slurm scripts, configuration and env setup files which are needed to run the end-to-end pipeline. Currently AWS environment is supported by default.
0.2.0
Removed dependency between pipelines and end-to-end batch continuum imaging pipeline.
End-to-End batch continuum imaging pipeline now functions as a tool which runs pipeline(s) as a single job. Dependency management delegated to the upstream process.
CLI Contract: End-to-End batch continuum imaging pipeline accepts a configuration file/stream and type of the configuration
Use
srunfor DASK cluster creation instead of dask-jobqueues in-order to use the resources allocated to the slurm job.Add support for monitoring using ska-sdp-benchmark-monitor
0.1.0
Initial release of the SKA SDP end-to-end batch continuum imaging pipeline