Releases: icgc-argo/workflow-management
Releases · icgc-argo/workflow-management
3.2.0
- VolumeMounts are now computed and passed to Nextflow so that only the required volumes are mounted in K8s
- Configurable plugins directory to support Nextflow 21+
- Configurable image pull policy for Nextflow pods (defaults to
IfNotPresent
) - Adding a thread sleep for
NextflowMonitor
to improve performance - Auth removed and
AuthDisabled
as default given Workflow-API/RabbitMQ integration
Package Updates
- Nextflow to
21.04.3
- Spring Boot to
2.5.4
- Reactor RabbitMQ Streams to
0.9.0
- Clean-up, update, and removal of supporting libs
3.0.0
- Management now comes with optional gatekeeper profile. When turned on, it records runs and determines whether/how runs are allowed to change state.
- Runs are now started or cancelled by passing messages via a RabbitMQ queue
- WES api and GQL for starting and cancelling runs are removed
- Jenkins updated to pass home directory for Docker builds
2.11.0
- Run Request updated to latest WES spec
- WorkflowEngineParams
latest
param type changed toBoolean
, previouslyString
- Nextflow workflow runs can be scheduled on any namespace (via appConfig), previously it could only use the same namespace as the management service
- Web security added to static paths
- Update java version (
openjdk11:jre-11.0.8_10-alpine
) in base container to avoid Bugs with K8s >1.17, thanks @bnwlf! - Update Jenkins to use Github repos
2.8.0
- Nextflow version 20.07.1
- GraphQL mutation exposed, federated in
rdpc-gateway
thereby allowing 100% GQL interaction with all WES services - New run profiles:
apiKey
- injects a kubernetes secret into workflow/process pods that can be access by the workflow (alternate way of providing api-key to running workflows for interacting with SONG/SCOREoauth2Token
- will be used for upcoming application tokens mechanism, allowing management to provision tokens on demand for workflowssecure | !secure
- endpoint (including graphQL) security based on JWT scopes
- Ability to "cancel" runs that have pods in
FAILED
state that would previously be considered stuck - Chart version bump for new values
- Refactor of run monitor system
2.1.1
2.1.0
- better error handling for nextflow workflows that do not make it to weblog
workflow_engine_params
andworkflow_params
are now completely optional (previously required empty object in request)workflow_engine_params.resume
must be in UUID format or else 400 error will occur- minor cosmetic tweaks and comments