- Update to version 0.0+git20200220.d3be16a:
* use pull mode instead of event mode to monitor the stream changes on broker, since original event mode will block the read async thread
* mkdir .cache if not exists
- Update to version 0.0+git20200219.7534b05:
* add script to install kube-prometheus and k8s-prometheus-adapter
* add checksum for download files
- Update to version 0.0+git20200219.4dab74d:
* add HPA yaml files
* fix python warning for unused return value
* when a task meet with unrecovery exception, then stop all tasks and stop application, let kubernete restart container for full recovery
* add initContainer for service dependency
* add more references for custom metrics based HPA
* add custom metrics based HPA
* add note for timezone setting
- Update to version 0.0+git20200218.db6ecc1:
* fix k8s document typo
* add HPA example yaml file on metric CPU usage
* fix broken link of doc reference
* add step for exposing NodePort for metrics previewing
* cleanup unused document reference
* add reference of metrics and scale in top README.md
* add document for dynamic scale and HPA
* add doc inference_metrics.md for detail explains on metric collection and monitoring
* add equation picture
* add equation picture
* add description of inference metrics monitoring
* add description of tested kubernetes version
* Correct the inference type with INT8/FP32 in dockerfile and standalone running scripts.
* Correct the inference type with INT8/FP32 in dockerfile and standalone running scripts.
* Add reference link for INT8 and FP32 model seperately.
* add whitepaper reference
- Update to version 0.0+git20200215.2b8765e:
* clean up the script of download-models.sh
* fix the bug that websocket could not get event of stream adding/remove
* fix the bug that websocket could not get event of stream adding/remove
* add int8 inference support
* use different macro for the path and name of inference model to avoid triggering the model download logic from openvino base container, which lead some issues for deployment
* refine the gen-k8s-yaml script to handle the special characters like : / in prefix(namespace) of registry name
* add pip mirror arg to accelerate docker build speed
* check whether npm is installed before run npm command
* Add customized grafana dashboard
* Update metric scale_ratio algorithm
* Update monitor docs
* HPA on prometheus metrics
* Add ei_scale_ratio metric for HPA
* Reset the infer and drop fps metrics when idle 30s
* Clean up the Dockerfile and script
* Move drop fps to the inference stage
* Use kube-prometheus for monitoring
- Use MYRIAD device instead of CPU
- Next round of adjustments for kustomize
- More changes to k8s-yaml samples for kustomize
- Rework k8s-yaml to fit better with kustomize
- Use official openSUSE redis image
- Rename dldt-python3 to python3-dldt
- Fix building on CODE15
- Add systemd service to run the css container
- run-css.patch: Adjust shell script to run the camera stream
service
- Add elastic-inference-camera-stream sub packages allowing
streaming a webcam into the cluster
- Fix dependencies of ei-css
- Adjust colors of stream view
- frame.py-error-handling.patch: do some error handling if the
redis service is not yet reacheable
- websocket-server-retry.patch: handle timeout if gateway server
is not coming up fast enough
- fix-person-detection-model-name.patch: use correct model for
person detection
- Use ogv files instead of mp4 for the samples
- Add conflicts to the ei-* packages as the are not parallel
installable
- Revert one user only change
- Create a k8s-yaml package
- Consolidate to one user only
- Enable health-check for redis calls
- Add ei-camera-stream-service
- Build models directory from git
- Add LICENSE files for sample videos and models
- Add ei-gateway-server
- Add ei-file-stream-service
- Add ei-websocket-server