v3.0.0

Release Notes for HySDS framework version 3.0.0 (Black Mamba)

Github release: https://github.com/hysds/hysds-framework/releases/tag/v3.0.0

Summary

HySDS framework (core) v3.0.0 includes bug fixes and major enhancements.

Highlights

• Port from Python 2.7 to Python 3.7+

  • Migration from using OS's built-in python distribution to an Anaconda-based python distribution

• Docker image loading enhancement using Docker registry and S3 backend

• Resiliency improvements when running at high load (NISAR Death Valley workload @ 3000 concurrent workers)

  • Optimizing System Resources and HySDS for Death Valley conditions

• Support for verdi worker execution of GPU job types

• Added product checksum validation pre-processor

• Improvements to job-spec and hysds-io specifications

  • JSON schema validation

  • job-spec fields now required: disk_usage, soft_time_limit, time_limit

• Migration to AWS autoscaling fleet to support multiple instance types and purchase options

• Validation of support for Matlab-based job types

• Various bug fixes in triage post-processor

• Speedup verdi, pge-base, and cuda-pge-base docker images' entrypoint when executing on overlay2 docker storage backend

• added optional capability to dump docker stats (e.g. CPU and memory usage) after execution of a job type's docker container

  • info on metrics being collected: https://www.datadoghq.com/blog/how-to-collect-docker-metrics/

• updates to metrics Kibana dashboard for Job Metrics

• created separate metrics and visualizations

  • job execution (pre-proprocessing steps + docker image loading + container execution + post-processing steps)

  • container execution only (PGE)

• added metrics and visualizations for container executions

  • total CPU usage (cumulative CPU usage including user, system, idle, etc.)

  • max memory usage

• How to enable docker execution stats for a job type

• PCM resource usage monitoring/alerting using CloudWatch

All updates