Backward compatability Issue with ModelName and Level for custom metrics

[agunapal]

🐛 Describe the bug

With 0.6.0 TorchServe, we observe the following for custom metrics

2023-11-07T21:34:24,676 [INFO ] W-9028-mnist_1.0-stdout MODEL_METRICS - ExtraInitializationTime.Milliseconds:1001.04|#ModelName:mnist,Level:Model|#hostname:ip-172-31-60-100,requestID:5348d06f-3c30-4ac0-b412-8b62b1e93a3a,timestamp:1699392864

2023-11-07T21:34:35,041 [INFO ] W-9003-mnist_1.0 org.pytorch.serve.wlm.WorkerThread - Flushing req. to backend at: 1699392875041
2023-11-07T21:34:35,042 [INFO ] W-9003-mnist_1.0-stdout MODEL_LOG - Backend received inference at: 1699392875
2023-11-07T21:34:35,053 [INFO ] W-9003-mnist_1.0-stdout MODEL_METRICS - batchSize.request:1|#ModelName:mnist,Level:Model|#hostname:ip-172-31-60-100,requestID:514b4052-0067-411b-8262-cdd127e6a83e,timestamp:1699392875

With 0.9.0 TorchServe, we observe the following

2023-10-27T10:43:07,174 - ExtraInitializationTime.ms:1000.18|#MethodName:initialize,Level:Error|#hostname:88665a372f4b.ant.amazon.com,requestID:null,timestamp:1698428587
2023-10-27T10:43:10,438 - batchSize.request:1.0|#HandlerName:MNISTDigitClassifier,MethodName:preprocess,ModelName:mnist,Level:Model|#hostname:88665a372f4b.ant.amazon.com,requestID:cd84d644-cec7-40b1-b236-48589c345d94,timestamp:1698428590

Error logs

Posted above

Installation instructions

No

Model Packaing

N/A

config.properties

No response

Versions

0.9.0

Repro instructions

N/A

Possible Solution

No response

cc @namannandan

[namannandan]

In versions up to 0.6.0, the request_id for model load requests are generated at random. Therefore, metrics updated in the initialize function of the handler have both ModelName and Level dimensions

serve/ts/model_loader.py

Lines 90 to 91 in 4236a86

	# TODO: Request ID is not given. UUID is a temp UUID.
	metrics = MetricsStore(uuid.uuid4(), model_name)

In versions 0.6.1 onward, the request_ids is not set, therefore, metrics updated in the initialize function only have the Level dimension set to Level:Error.

serve/ts/model_service_worker.py

Lines 65 to 67 in 0510772

	self.metrics_cache = MetricsCacheYamlImpl(config_file_path=metrics_config)
	if self.metrics_cache:
	self.metrics_cache.initialize_cache()