[Feat] Add Sentry Continuous Profiling Support to vLLM Router (vllm-project#624)

* Added profile lifecycle tracking and sampling rates for traces and sessions in Sentry initialization.

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

* Added validation and command-line arguments for Sentry traces and profile session sample rates in the parser. Ensured sample rates are within the range of 0.0 to 1.0.

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

* Update README.md to include new command-line options for Sentry traces and profiling session sample rates.

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

* Refactor error message formatting in validate_args function for Sentry profile session sample rate validation.

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

* Fix Dockerfile syntax by changing 'source' to '.' for activating the virtual environment.

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

---------

Signed-off-by: Ifta Khairul Alam Adil <ikaadil007@gmail.com>

Author: ikaadil

docker/Dockerfile.kvaware

@@ -19,7 +19,7 @@ ARG INSTALL_OPTIONAL_DEP=semantic_cache,lmcache
 ENV INSTALL_OPTIONAL_DEP=${INSTALL_OPTIONAL_DEP}
 
 # hadolint ignore=SC1091
-RUN source /opt/venv/bin/activate && \
+RUN . /opt/venv/bin/activate && \
     uv pip install --upgrade --no-cache-dir pip setuptools_scm && \
     uv pip install --no-cache-dir .
 

src/vllm_router/README.md

@@ -54,6 +54,8 @@ The router can be configured using command-line arguments. Below are the availab
 ### Sentry Options
 
 - `--sentry-dsn`: The Sentry Data Source Name to use for error reporting.
+- `--sentry-traces-sample-rate`: The sample rate for Sentry traces (0.0 to 1.0). Default is 0.1 (10%).
+- `--sentry-profile-session-sample-rate`: The sample rate for Sentry profiling sessions (0.0 to 1.0). Default is 1.0 (100%).
 
 ## Build docker image
 

src/vllm_router/app.py

@@ -116,7 +116,13 @@ def initialize_all(app: FastAPI, args):
         ValueError: if the service discovery type is invalid
     """
     if sentry_dsn := args.sentry_dsn:
-        sentry_sdk.init(dsn=sentry_dsn, send_default_pii=True)
+        sentry_sdk.init(
+            dsn=sentry_dsn,
+            send_default_pii=True,
+            profile_lifecycle="trace",
+            traces_sample_rate=args.sentry_traces_sample_rate,
+            profile_session_sample_rate=args.sentry_profile_session_sample_rate,
+        )
 
     if args.service_discovery == "static":
         initialize_service_discovery(

src/vllm_router/parsers/parser.py

@@ -91,6 +91,12 @@ def validate_args(args):
         raise ValueError("Engine stats interval must be greater than 0.")
     if args.request_stats_window <= 0:
         raise ValueError("Request stats window must be greater than 0.")
+    if not (0.0 <= args.sentry_traces_sample_rate <= 1.0):
+        raise ValueError("Sentry traces sample rate must be between 0.0 and 1.0.")
+    if not (0.0 <= args.sentry_profile_session_sample_rate <= 1.0):
+        raise ValueError(
+            "Sentry profile session sample rate must be between 0.0 and 1.0."
+        )
 
 
 def parse_args():
@@ -298,6 +304,20 @@ def parse_args():
         help="Enables Sentry Error Reporting to the specified Data Source Name",
     )
 
+    parser.add_argument(
+        "--sentry-traces-sample-rate",
+        type=float,
+        default=0.1,
+        help="The sample rate for Sentry traces. Default is 0.1 (10%)",
+    )
+
+    parser.add_argument(
+        "--sentry-profile-session-sample-rate",
+        type=float,
+        default=1.0,
+        help="The sample rate for Sentry profiling sessions. Default is 1.0 (100%)",
+    )
+
     parser.add_argument(
         "--prefill-model-labels",
         type=str,