0.9.3 (2024-03-04)
- certificates: update expired CERN Grid CA certificate (#440) (8d6539a), closes #439
- docker: non-editable submodules in "latest" mode (#416) (3bdda63)
- python: bump all required packages as of 2024-03-04 (#442) (de119eb)
- python: bump shared REANA packages as of 2024-03-04 (#442) (fc77628)
- db: set job status also in the main database (#423) (9d6fc99)
- docs: move from reST to Markdown (#428) (4732884)
- monitor: centralise logs and status updates (#423) (3685b01)
- monitor: move fetching of logs to job-manager (#423) (1fc117e)
- commitlint: addition of commit message linter (#417) (f547d3b)
- commitlint: allow release commit style (#443) (0fc9794)
- commitlint: check for the presence of concrete PR number (#425) (35bc1c5)
- pytest: move to PostgreSQL 14.10 (#429) (42622fa)
- release-please: initial configuration (#417) (fca6f74)
- release-please: update version in Dockerfile/OpenAPI specs (#421) (e6742f2)
- shellcheck: fix exit code propagation (#425) (8e74a85)
- Adds metadata labels to Dockerfile.
- Adds automated multi-platform container image building for amd64 and arm64 architectures.
- Changes CVMFS support to allow users to automatically mount any available repository.
- Fixes container image building on the arm64 architecture.
- Fixes the creation of Kubernetes jobs by retrying in case of error and by correctly handling the error after reaching the retry limit.
- Fixes job monitoring in cases when job creation fails, for example when it is not possible to successfully mount volumes.
- Adds unique error messages to Kubernetes job monitor to more easily identify source of problems.
- Changes Paramiko to version 3.0.0.
- Changes HTCondor to version 9.0.17 (LTS).
- Changes Rucio authentication helper to version 1.1.1 allowing users to override the Rucio server and authentication hosts independently of VO name.
- Fixes intermittent Slurm connection issues by DNS-resolving the Slurm head node IPv4 address before establishing connections.
- Fixes deletion of failed jobs not being performed when Kerberos is enabled.
- Fixes job monitoring to consider OOM-killed jobs as failed.
- Fixes Slurm command generation issues when using fully-qualified image names.
- Fixes location of HTCondor build dependencies.
- Fixes detection of default Rucio server and authentication host for ATLAS VO.
- Fixes container image names to be Podman-compatible.
- Adds support for Rucio authentication for workflow jobs.
- Adds support for specifying
slurm_partition
andslurm_time
for Slurm compute backend jobs. - Adds Kerberos sidecar container to renew ticket periodically for long-running jobs.
- Changes
reana-auth-vomsproxy
sidecar to the latest stable version to support client-side proxy file generation technique and ESCAPE VOMS. - Changes default Slurm partition to
inf-short
. - Changes to PostgreSQL 12.13.
- Changes the base image of the component to Ubuntu 20.04 LTS and reduces final Docker image size by removing build-time dependencies.
- Adds support for specifying
kubernetes_job_timeout
for Kubernetes compute backend jobs. - Adds a new condition to allow processing jobs in case of receiving multiple failed events when job containers are not in a running state.
- Adds database connection closure after each REST API request.
- Adds labels to job and run-batch pods to reduce k8s events to listen to for
job-monitor
. - Fixes auto-mounting of Kubernetes API token inside user jobs by disabling it.
- Changes job dispatching to use only job-specific node labels.
- Changes to PostgreSQL 12.8.
- Changes HTCondor to 8.9.11.
- Changes myschedd package and configuration to latest versions.
- Fixes job command formatting bug for CWL workflows on HTCondor.
- Adds configuration environment variable to set job memory limits for the Kubernetes compute backend (
REANA_KUBERNETES_JOBS_MEMORY_LIMIT
). - Fixes Kubernetes job log capture to include information about failures caused by external factors such as OOMKilled.
- Adds support for specifying
kubernetes_memory_limit
for Kubernetes compute backend jobs.
- Adds new configuration to toggle Kubernetes user jobs clean up.
- Fixes HTCondor Docker networking and machine version requirement setup.
- Fixes HTCondor logs and workspace files retrieval on job failure.
- Fixes Slurm job submission providing the correct shell environment to run Singularity.
- Changes HTCondor myschedd to the latest version.
- Changes job status
succeeded
tofinished
to use central REANA nomenclature. - Changes how to deserialise job commands using central REANA-Commons deserialiser function.
- Fixes minor code warnings.
- Changes CI system to include Python flake8 and Dockerfile hadolint checkers.
- Adds support for specifying
htcondor_max_runtime
andhtcondor_accounting_group
for HTCondor compute backend jobs. - Fixes Docker build by properly exiting when there are problems with
myschedd
installation.
- Adds support for running unpacked Docker images from CVMFS on HTCondor jobs.
- Adds support for pulling private images using image pull secrets.
- Adds support for VOMS proxy as a new authentication method.
- Adds pinning of all Python dependencies allowing to easily rebuild component images at later times.
- Fixes HTCondor job submission retry technique.
- Changes error reporting on Docker image related failures.
- Changes runtime pods to prefix user workflows with the configured REANA prefix.
- Changes CVMFS to be read-only mount.
- Changes runtime job instantiation into the configured runtime namespace.
- Changes test suite to enable running tests locally also on macOS platform.
- Changes CERN HTCondor compute backend to use the new
myschedd
connection library. - Changes CERN Slurm compute backend to improve job status detection.
- Changes base image to use Python 3.8.
- Changes code formatting to respect
black
coding style. - Changes documentation to single-page layout.
- Upgrades REANA-Commons package using latest Kubernetes Python client version.
- Adds generic job manager class and provides example classes for CERN HTCondor and CERN Slurm clusters.
- Moves job controller to the same Kubernetes pod with the REANA-Workflow-Engine-* (sidecar pattern).
- Adds sidecar container to the Kubernetes job pod if Kerberos authentication is required.
- Provides user secrets to the job container runtime tasks.
- Refactors job monitoring using singleton pattern.
- Pins
urllib3
due to a conflict while installingKubernetes
Python library. - Fixes documenation build badge.
- Adds a new endpoint to delete jobs (Kubernetes).
- Introduces new common interface for job management which defines what the compute backends should offer to be compatible with REANA, currently only Kubernetes backend is supported.
- Fixes security vulnerability which allowed users to access other people's workspaces.
- Makes CVMFS mounts optional and configurable at repository level.
- Updates the creation of CVMFS volumes specification, it now uses normal persistent volume claims.
- Increases stability and improves test coverage.
- Improves REST API documentation rendering.
- Changes license to MIT.
- Adapts Kubernetes API adaptor to mount shared volumes on jobs as CEPH
persistentVolumeClaim
's (managed byreana-cluster
) instead of plain CEPH volumes.
- Pins REANA-Commons and REANA-DB dependencies.
- Adds uwsgi for production deployments.
- Switches from pykube to official Kubernetes python client.
- Adds compatibility with latest Kubernetes.
- Adds dockerignore file to ease developments.
- Initial public release.