Ensure Model Running

Author: hi4rahul

functions/start_model/app.py

@@ -25,10 +25,10 @@ def lambda_handler(event, context):
     rekog_client = boto3.client('rekognition')
     projectversionarn = os.environ['rekog_model_project_version_arn']
     projectarn = os.environ['rekog_model_project_arn']
-    running_states = ['STARTING', 'RUNNING']
+    # running_states = ['STARTING', 'RUNNING']
     projectversionname = projectversionarn.split("/")[3]
     # Check if already running
-    # Call Custom Rekog
+    # Call Custom Rekognition project version
     try:
         isrunning_response = rekog_client.describe_project_versions(
             ProjectArn=projectarn,
@@ -38,17 +38,22 @@ def lambda_handler(event, context):
         print(e)
 
     running_status = isrunning_response['ProjectVersionDescriptions'][0]['Status']
-    if running_status in running_states:
+    if running_status == 'RUNNING':
         # Do nothing
         print('Model Start Status: %s' % running_status)
+        return 'RUNNING'
+    if running_status == 'STARTING':
+        # Do nothing
+        print('Model Start Status: %s' % running_status)
+        return 'STARTING'
     else:
         # If not running - Start
         try:
             running_status = rekog_client.start_project_version(
-                ProjectVersionArn=projectversionarn,                
-                MinInferenceUnits=1 #Can be increased upto 5 for running multiple inference units
+                ProjectVersionArn = projectversionarn,                
+                MinInferenceUnits = 1 #Can be increased upto 5 for running multiple inference units
             )
         except Exception as e:
             print(e)
-        
-    return running_status
+        return running_status
+

template.yaml

@@ -12,7 +12,7 @@ Parameters:
 Mappings:
   Solution:
     Constants:
-      Version: 'v0.4'
+      Version: 'v0.5'
 
 Resources:
   SourceS3Bucket:
@@ -100,7 +100,7 @@ Resources:
           Wait for the model to start:
             Type: Wait
             Seconds: 600
-            Next: Enable SQS Trigger
+            Next: Start Model
           Keep Model running for 1 hr:
             Type: Wait
             Seconds: 3540