Expose kernel_info_timeout configuration

docs/source/operators/config-add-env.md

@@ -32,6 +32,9 @@ Besides those environment variables associated with configurable options, the fo
     startup attempt will take place.  If a second timeout occurs, Enterprise
     Gateway will report a failure to the client.
 
+  EG_KERNEL_LAUNCH_TIMEOUT=60
+    The time (in seconds) Enterprise Gateway will wait before giving up on a kernel
+
   EG_SENSITIVE_ENV_KEYS=""
     A comma separated list (e.g. "secret,pwd,auth") of sensitive environment
     variables. Any environment variables that contain any of the words from this

docs/source/users/kernel-envs.md

@@ -54,6 +54,9 @@ There are several supported `KERNEL_` variables that the Enterprise Gateway serv
     be submitted in the kernel startup if that particular kernel's startup time is
     expected to exceed that of the EG_KERNEL_LAUNCH_TIMEOUT set when Enterprise
     Gateway starts.
+    
+  KERNEL_INFO_TIMEOUT=<from user> or EG_KERNEL_INFO_TIMEOUT=60
+    The time (in seconds) Enterprise Gateway will wait before giving up on a kernel
 
   KERNEL_NAMESPACE=<from user> or KERNEL_POD_NAME or EG_NAMESPACE
     Kubernetes only.  This indicates the name of the namespace to use or create on

enterprise_gateway/services/kernels/remotemanager.py

@@ -26,6 +26,7 @@
 from ..sessions.kernelsessionmanager import KernelSessionManager
 
 default_kernel_launch_timeout = float(os.getenv("EG_KERNEL_LAUNCH_TIMEOUT", "30"))
+default_kernel_info_timeout = float(os.getenv("EG_KERNEL_INFO_TIMEOUT", "60"))
 kernel_restart_status_poll_interval = float(os.getenv("EG_RESTART_STATUS_POLL_INTERVAL", 1.0))
 
 
@@ -437,6 +438,7 @@ def __init__(self, **kwargs: dict[str, Any] | None):
         self.kernel_id = None
         self.user_overrides = {}
         self.kernel_launch_timeout = default_kernel_launch_timeout
+        self.kernel_info_timeout = default_kernel_info_timeout
         self.restarting = False  # need to track whether we're in a restart situation or not
         self._activity_stream = None
 
@@ -513,6 +515,10 @@ def _capture_user_overrides(self, **kwargs: dict[str, Any] | None) -> None:
         self.kernel_launch_timeout = float(
             env.get("KERNEL_LAUNCH_TIMEOUT", default_kernel_launch_timeout)
         )
+        # if KERNEL_INFO_TIMEOUT is passed in the payload, override it.
+        self.kernel_launch_timeout = float(
+            env.get("KERNEL_INFO_TIMEOUT", default_kernel_info_timeout)
+        )
         self.user_overrides.update(
             {
                 key: value

enterprise_gateway/services/processproxies/processproxy.py

@@ -61,6 +61,7 @@
 redaction_mask = os.getenv("EG_REDACTION_MASK", "********")
 
 default_kernel_launch_timeout = float(os.getenv("EG_KERNEL_LAUNCH_TIMEOUT", "30"))
+default_kernel_info_timeout = float(os.getenv("EG_KERNEL_INFO_TIMEOUT", "60"))
 max_poll_attempts = int(os.getenv("EG_MAX_POLL_ATTEMPTS", "10"))
 poll_interval = float(os.getenv("EG_POLL_INTERVAL", "0.5"))
 socket_timeout = float(os.getenv("EG_SOCKET_TIMEOUT", "0.005"))
@@ -430,6 +431,7 @@ def __init__(self, kernel_manager: RemoteKernelManager, proxy_config: dict):  #
 
         self.kernel_id = self.kernel_manager.kernel_id
         self.kernel_launch_timeout = default_kernel_launch_timeout
+        self.kernel_info_timeout = default_kernel_info_timeout
         self.lower_port = 0
         self.upper_port = 0
         self._validate_port_range()

etc/docker/docker-compose.yml

@@ -21,6 +21,7 @@ services:
     environment:
       - "EG_DOCKER_NETWORK=${EG_DOCKER_NETWORK:-enterprise-gateway_enterprise-gateway}"
       - "EG_KERNEL_LAUNCH_TIMEOUT=${EG_KERNEL_LAUNCH_TIMEOUT:-60}"
+      - "EG_KERNEL_INFO_TIMEOUT=${EG_KERNEL_INFO_TIMEOUT:-60}"
       - "EG_CULL_IDLE_TIMEOUT=${EG_CULL_IDLE_TIMEOUT:-3600}"
       # Use double-defaulting for B/C.  Support for EG_KERNEL_WHITELIST will be removed in a future release
       - "EG_ALLOWED_KERNELS=${EG_ALLOWED_KERNELS:-${EG_KERNEL_WHITELIST:-'r_docker','python_docker','python_tf_docker','python_tf_gpu_docker','scala_docker'}}"

etc/docker/enterprise-gateway/start-enterprise-gateway.sh

@@ -29,6 +29,7 @@ export EG_CULL_CONNECTED=${EG_CULL_CONNECTED:-False}
 EG_ALLOWED_KERNELS=${EG_ALLOWED_KERNELS:-${EG_KERNEL_WHITELIST:-"null"}}
 export EG_ALLOWED_KERNELS=`echo ${EG_ALLOWED_KERNELS} | sed 's/[][]//g'` # sed is used to strip off surrounding brackets as they should no longer be included.
 export EG_DEFAULT_KERNEL_NAME=${EG_DEFAULT_KERNEL_NAME:-python_docker}
+export EG_KERNEL_INFO_TIMEOUT=${EG_KERNEL_INFO_TIMEOUT:-60}
 
 # Determine whether the kernels-allowed list should be added to the start command.
 # This is conveyed via a 'null' value for the env - which indicates no kernel names
@@ -46,4 +47,5 @@ exec jupyter enterprisegateway \
 	--RemoteMappingKernelManager.cull_idle_timeout=${EG_CULL_IDLE_TIMEOUT} \
 	--RemoteMappingKernelManager.cull_interval=${EG_CULL_INTERVAL} \
 	--RemoteMappingKernelManager.cull_connected=${EG_CULL_CONNECTED} \
-	--RemoteMappingKernelManager.default_kernel_name=${EG_DEFAULT_KERNEL_NAME}
+	--RemoteMappingKernelManager.default_kernel_name=${EG_DEFAULT_KERNEL_NAME} \
+	--RemoteMappingKernelManager.kernel_info_timeout=${EG_KERNEL_INFO_TIMEOUT}

etc/kubernetes/helm/enterprise-gateway/templates/deployment.yaml

@@ -71,6 +71,8 @@ spec:
           value: {{ .Values.logLevel }}
         - name: EG_KERNEL_LAUNCH_TIMEOUT
           value: !!str {{ .Values.kernel.launchTimeout }}
+        - name: EG_KERNEL_INFO_TIMEOUT
+          value: !!str {{ .Values.kernel.infoTimeout }}
         - name: EG_ALLOWED_KERNELS
           value: {{ toJson .Values.kernel.allowedKernels | squote }}
         - name: EG_DEFAULT_KERNEL_NAME