Redshift hpc

sklearn/stock-pred-redshift/model/evaluation.py

@@ -12,6 +12,8 @@
 r_database = os.getenv('DKUBE_DATASET_REDSHIFT_DATABASE', None)
 r_user = os.getenv('DKUBE_DATASET_REDSHIFT_USER', None)
 
+dkube_url = os.getenv('DKUBE_CONTROLLER_NODEIP_ACCESS_URL', "http://dkube-controller-worker.dkube:5000")
+
 if 'https:' in r_endpoint:
     p = '(?:https*://)?(?P<host>[^:/ ]+).?(?P<port>[0-9]*).*'
     m = re.search(p,r_endpoint)
@@ -28,6 +30,9 @@
 degree= int(sys.argv[5]) if len(sys.argv) > 5 else 2
 
 def log_metrics(key, value):
+    #TODO - reporting has to be changed to mlflow api or expose dkube-exporter via gateway
+    if MLFLOW_METRIC_REPORTING != "True":
+        return
     url = "http://dkube-exporter.dkube:9401/mlflow-exporter"
     train_metrics = {}
     train_metrics['mode']="train"
@@ -48,7 +53,7 @@ def eval_metrics(actual, pred):
 
 def rs_fetch_datasets():
     user = os.getenv("DKUBE_USER_LOGIN_NAME")
-    url = "http://dkube-controller-worker.dkube:5000/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s"
+    url = dkube_url + "/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s"
     headers={"authorization": "Bearer "+os.getenv("DKUBE_USER_ACCESS_TOKEN")}
     datasets = []
     for ds in json.load(open('/etc/dkube/redshift.json')):
@@ -116,4 +121,4 @@ def get_data(filename):
     log_metrics('R2', r2)
     print('RMSE', rmse)
     print('R2', r2)
-    print('MAE', mae)
+    print('MAE', mae)

sklearn/stock-pred-redshift/model/printRecords.R

@@ -5,7 +5,8 @@ ds <- fromJSON("/etc/dkube/redshift.json")
 
 user <- Sys.getenv("LOGNAME")
 
-url <- "http://dkube-controller-worker.dkube:5000/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s"
+dkube_url <- Sys.getenv("DKUBE_CONTROLLER_NODEIP_ACCESS_URL")
+url <- sprintf("%s/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s",dkube_url)
 
 token <- strsplit(Sys.getenv("RSTUDIO_HTTP_REFERER"), "=")[[1]][2]
 
@@ -29,4 +30,4 @@ get_password <- function(user, db){
   }
 }
 
-get_password("dpaks", "dkube")
+get_password("dpaks", "dkube")

sklearn/stock-pred-redshift/model/train.py

@@ -9,10 +9,12 @@
 import joblib, psycopg2
 import requests
 
+MLFLOW_METRIC_REPORTING = os.getenv('MLFLOW_METRIC_REPORTING', "True")
 
 r_endpoint = os.getenv('DKUBE_DATASET_REDSHIFT_ENDPOINT', None)
 r_database = os.getenv('DKUBE_DATASET_REDSHIFT_DATABASE', None)
 r_user = os.getenv('DKUBE_DATASET_REDSHIFT_USER', None)
+dkube_url = os.getenv('DKUBE_CONTROLLER_NODEIP_ACCESS_URL', "http://dkube-controller-worker.dkube:5000")
 
 
 if 'https:' in r_endpoint:
@@ -39,11 +41,11 @@ def eval_metrics(actual, pred):
 
 def rs_fetch_datasets():
     user = os.getenv("DKUBE_USER_LOGIN_NAME")
-    url = "http://dkube-controller-worker.dkube:5000/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s"
+    url = dkube_url + "/dkube/v2/controller/users/%s/datums/class/dataset/datum/%s"
     headers={"authorization": "Bearer "+os.getenv("DKUBE_USER_ACCESS_TOKEN")}
     datasets = []
     for ds in json.load(open('/etc/dkube/redshift.json')):
-        resp = requests.get(url % (user, ds.get('rs_name')), headers=headers).json()
+        resp = requests.get(url % (user, ds.get('rs_name')), headers=headers, verify=False).json()
         ds['rs_password'] = resp['data']['datum']['redshift']['password']
         datasets.append(ds)
     return datasets
@@ -58,6 +60,9 @@ def get_password(user, db):
 MODEL_DIR = '/opt/dkube/model'
 
 def log_metrics(key, value):
+    #TODO - reporting has to be changed to mlflow api or expose dkube-exporter via gateway
+    if MLFLOW_METRIC_REPORTING != "True":
+        return
     url = "http://dkube-exporter.dkube:9401/mlflow-exporter"
     train_metrics = {}
     train_metrics['mode']="train"
@@ -145,4 +150,4 @@ def get_data(filename):
 
     img = cv2.imread('svm.png')
     log_histogram('Stock Prices', prices, step=1)
-    log_images('Stock Predictions Graph',[img])
+    log_images('Stock Predictions Graph',[img])