Improvements and Fixes: Automation (Samagra-Development#256)

* Ability to add deployment constraints as a part of config.json

---------

Co-authored-by: Gautam <[email protected]>

Author: singhalkarun

config.json

@@ -22,7 +22,8 @@
       "apiBasePath": "text_classification/flow_classification/local/",
       "containerPort": 8000,
       "environment": {},
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.role==worker"]
     },
     {
       "serviceName": "text_translation_azure_dict",
@@ -42,23 +43,26 @@
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },    
     {
       "serviceName": "coref_spacy",
       "modelBasePath": "src/coref/spacy/local/.",
       "apiBasePath": "/coref/spacy/local",
       "containerPort": 8000,
       "environment": {},
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "coref_bart",
       "modelBasePath": "src/coref/bart/local/.",
       "apiBasePath": "/coref/bart/local",
       "containerPort": 8000,
       "environment": {},
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     }, 
      {
       "serviceName": "text_translation_azure",
@@ -68,45 +72,51 @@
       "environment": {
         "AZURE_TRANSLATE_KEY": "${AZURE_TRANSLATE_KEY}"
       },
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "asr_mms",
       "modelBasePath": "src/asr/fairseq_mms/local/.",
       "apiBasePath": "/asr/fairseq_mms/local",
       "containerPort": 8000,
       "environment": {},
-      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"]
+      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "coref_fcoref",
       "modelBasePath": "src/coref/fcoref/local/.",
       "apiBasePath": "/coref/fcoref/local",
       "containerPort": 8000,
       "environment": {},
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "text_translation_bhashini",
       "modelBasePath": "src/text_translation/bhashini/remote/.",
       "apiBasePath": "/text_translation/bhashini/remote",
       "containerPort": 8000,
       "environment": {},
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "text_translation_ai4bharat",
       "modelBasePath": "src/text_translation/ai4bharat/remote/.",
       "apiBasePath": "src/text_translation/ai4bharat/remote",
       "containerPort": 8000,
-      "environment": {}
+      "environment": {},
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "text_lang_detection_bhashini",
       "modelBasePath": "src/text_lang_detection/bhashini/remote/.",
       "apiBasePath": "/text_lang_detection/bhashini/remote",
       "containerPort": 8000,
-      "environment": {}
+      "environment": {},
+      "constraints": ["node.labels.node_vm_type==gpu"]
       },
     {
       "serviceName": "chunking_mpnet",
@@ -116,7 +126,8 @@
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
-      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"]
+      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "embedding_instructor",
@@ -126,7 +137,8 @@
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
-      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"]
+      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     },
     {
       "serviceName": "llm_openai_gpt3",
@@ -136,7 +148,8 @@
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
-      "nginx": []
+      "nginx": [],
+      "constraints": ["node.labels.node_vm_type==gpu"]
     }
   ]
 }

generate_independent_docker.sh

@@ -22,6 +22,8 @@ for ((i=0; i<$count; i++)); do
     apiBasePath=$(jq -r ".models[$i].apiBasePath" config.json)
     containerPort=$(jq -r ".models[$i].containerPort" config.json)
 
+    countConstraints=$(jq ".models[$i].constraints | length" config.json)
+
     # Calculate the exposed port for the model
     exposedPort=$((8000 + i))
 
@@ -31,6 +33,15 @@ for ((i=0; i<$count; i++)); do
     # Add service details to docker-compose.yaml
     printf "  ${serviceName}:\n    image: ${DOCKER_REGISTRY_URL}/${GITHUB_REPOSITORY_URL}/${serviceName}:latest\n    ports:\n      - ${exposedPort}:${containerPort}\n" >> docker-compose-independent-generated.yaml
 
+    if [[ countConstraints -gt 0 ]]; then
+        printf "    deploy:\n      placement:\n        constraints:\n" >> docker-compose-independent-generated.yaml
+    fi
+    for ((j=0; j<$countConstraints; j++)); do
+        constraintLine=$(jq -r ".models[$i].constraints[$j]" config.json)
+
+        printf "          - ${constraintLine}\n" >> docker-compose-independent-generated.yaml
+    done
+
     # Add environment variables to docker-compose.yaml
     if [[ ${#environment[@]} -gt 0 ]]; then
         printf "    environment:\n" >> docker-compose-independent-generated.yaml