It’s a high-performance architecture for ‘deep’ linguistic processing of documents with DELPHI-IN tools pyDelphin and ACE combined with a parallel processing using kubernetes and RQ.
It is important to ensure that you have IBM Cloud account, if you do not sign up here. You also need to instanciate IBM Cloud Kubernetes Service and IBM Cloud Object Storage. Create a IBM Cloud Object Storage service credential. Ensure that you have installed:
First step is log in your IBM Cloud account and then set the cluster you created as the context for this session:
ibmcloud login -a cloud.ibm.com -r us-south -g Default --sso
ibmcloud ks cluster-config --cluster $cluster_name
Installing Tiller, the Helm server, in your cluster:
kubectl create serviceaccount tiller -n kube-system kubectl create clusterrolebinding tiller --clusterrole=cluster-admin --serviceaccount=kube-system:tiller -n kube-system helm init --service-account tiller
Installing the IBM Cloud Object Storage plug-in:
helm repo add iks-charts https://icr.io/helm/iks-charts helm repo update helm fetch --untar iks-charts/ibm-object-storage-plugin helm plugin install ./ibm-object-storage-plugin/helm-ibmc chmod 755 ~/.helm/plugins/helm-ibmc/ibmc.sh helm ibmc install iks-charts/ibm-object-storage-plugin --name ibm-object-storage-plugin
Now fill the delphin-parsing-chart/values.yaml
file. Values decscription:
ace.options
ACE Command-line Optionsace.replicas
number of ace instancescredentials.apikey
the API key that you can retrieve from your IBM Cloud Object Storage service credentials earliercredentials.guid
the GUID of your IBM Cloud Object Storage service instance you can get that withibmcloud resource service-instance <service_name> | grep GUI
cos.inbucket
the name of the existing bucket with the data that you want to processcos.outbucket
the name of the existing bucket where the results will be savecos.endpoint
see https://cloud.ibm.com/docs/services/cloud-object-storage/basics?topic=cloud-object-storage-endpoints#select-regions-and-endpointscos.storageClassName
find your service instance in the IBM Cloud dashboard and review the Class and Location of your existing bucket. Then, use the appropriate storage class.
Finally, deploy:
helm install ./delphin-parsing-chart/
Get the public IP address for the worker node in the cluster.
ibmcloud ks workers --cluster $cluster_name
Open a browser and check out the app with the following URL: http://<IP_address>:30001
.
- Alexandre Rademaker
- Henrique Muniz
TODO