Open wuodar opened 2 months ago
Currently, one can set no-overlap to true, to avoid two concurrent executions. But is it possible to avoid overlapping between services? E.g.
no-overlap
version: '3.7' services: feature_pipeline: image: feature_pipeline container_name: feature_pipeline restart: always user: user build: context: . dockerfile: Dockerfile volumes: - ${DATA_VOLUME}:/home/user/app/data deploy: resources: limits: cpus: ${MAX_CPUS} memory: ${MAX_MEMORY} reservations: cpus: ${MIN_CPUS} memory: ${MIN_MEMORY} labels: ofelia.enabled: "true" ofelia.job-exec.feature_pipeline.schedule: ${PIPELINE_SCHEDULE} ofelia.job-exec.feature_pipeline.command: "python src/feature_pipeline.py" ofelia.job-exec.feature_pipeline.no-overlap: "true" inference_pipeline: image: inference_pipeline container_name: inference_pipeline restart: always user: user build: context: . dockerfile: Dockerfile volumes: - ${DATA_VOLUME}:/home/user/app/data deploy: resources: limits: cpus: ${MAX_CPUS} memory: ${MAX_MEMORY} reservations: cpus: ${MIN_CPUS} memory: ${MIN_MEMORY} labels: ofelia.enabled: "true" ofelia.job-exec.inference_pipeline.schedule: ${PIPELINE_SCHEDULE} ofelia.job-exec.inference_pipeline.command: "python src/inference_pipeline.py" ofelia.job-exec.inference_pipeline.no-overlap: "true" ofelia.job-exec.feature_pipeline.no-overlap: "true" ofelia: image: mcuadros/ofelia:latest restart: always depends_on: - feature_pipeline command: daemon --docker volumes: - /var/run/docker.sock:/var/run/docker.sock:ro
In this example, I'd like to set ofelia.job-exec.feature_pipeline.no-overlap: "true" but not only for feature_pipeline service, but also avoid running concurrently feature_pipeline and inference_pipeline.
ofelia.job-exec.feature_pipeline.no-overlap: "true"
feature_pipeline
inference_pipeline
No, that's not something Ofelia supports at the moment.
Currently, one can set
no-overlap
to true, to avoid two concurrent executions. But is it possible to avoid overlapping between services? E.g.In this example, I'd like to set
ofelia.job-exec.feature_pipeline.no-overlap: "true"
but not only forfeature_pipeline
service, but also avoid running concurrentlyfeature_pipeline
andinference_pipeline
.