feat(launch): add agent.startTimeout value for `WANDB_LAUNCH_START_…

…TIMEOUT` (#38)
wandb · Nov 1, 2023 · 552b7e8 · 552b7e8
1 parent 0e40471
commit 552b7e8
Show file tree

Hide file tree

Showing 4 changed files with 7 additions and 2 deletions.
diff --git a/charts/launch-agent/Chart.yaml b/charts/launch-agent/Chart.yaml
@@ -3,7 +3,7 @@ name: launch-agent
 icon: https://em-content.zobj.net/thumbs/240/apple/354/rocket_1f680.png
 description: A Helm chart for running the W&B Launch Agent in Kubernetes
 type: application
-version: 0.11.1
+version: 0.11.2
 maintainers:
   - name: wandb
     email: [email protected]

diff --git a/charts/launch-agent/README.md b/charts/launch-agent/README.md
@@ -45,7 +45,8 @@ The table below describes all the available variables in the chart:
 | `agent.useExternalWandbSecret` | bool            | **false**            | ""                              | Used to indicate you want to provide the api key secret external to this chart.                                                                  |
 | `agent.image`                  | string          | No                   | `wandb/launch-agent-dev:latest` | Container image for the agent.                                                                                                                   |
 | `agent.imagePullPolicy`        | string          | No                   | `Always`                        | Pull policy for the agent container image.                                                                                                       |
-| `agent.resources`              | object          | No                   | Limit to 1 CPU, 1Gi RAM         | Pod spec resources block for the agent.                                                                                                          |
+| `agent.resources`              | object          | No                   | Limit to 1 CPU, 1Gi RAM         | Pod spec resources block for the agent. true                                                                                                     |
+| `agent.startTimeout`           | int             | No                   | `1800`                          | Timeout in seconds that the agent will wait for a job to start before timing out.                                                                |
 | `namespace`                    | string          | No                   | `wandb`                         | The namespace to deploy the agent into.                                                                                                          |
 | `additionalTargetNamespaces`   | list(string)    | No                   | [`wandb`,`default`]             | A list of namespaces the agent can run jobs in.                                                                                                  |
 | `baseUrl`                      | string          | No                   | `https://api.wandb.ai`          | URL of your W&B server api.                                                                                                                      |

diff --git a/charts/launch-agent/templates/deployment.yaml b/charts/launch-agent/templates/deployment.yaml
@@ -53,6 +53,8 @@ spec:
                   key: wandb-base-url
             - name: CM_CHECKSUM
               value: {{ include (print $.Template.BasePath "/configmap.yaml") . | sha256sum }}
+            - name: WANDB_LAUNCH_START_TIMEOUT
+              value: "{{ .Values.agent.startTimeout }}"
           volumeMounts:
             - name: wandb-launch-config
               mountPath: /home/launch_agent/.config/wandb

diff --git a/charts/launch-agent/values.yaml b/charts/launch-agent/values.yaml
@@ -13,6 +13,8 @@ agent:
     limits:
       cpu: 1000m
       memory: 1Gi
+  # Job start timeout (seconds)
+  startTimeout: 1800
 
 # Namespace to deploy launch agent into
 namespace: wandb