[BUG] startupProbe failureThreshold missing from templates (Backend and Workflows/Workers)

[ndpete]

[BUG] startupProbe failureThreshold missing from templates (Backend and Workflows/Workers)

Description

The Retool Helm Chart (v6.10.3) contains two issues regarding startupProbe implementation that prevent users from configuring a sufficient safety window for long-running startups:

1. Omitted failureThreshold (Backend): The deployment_backend.yaml template does not include the failureThreshold key. Consequently, Kubernetes defaults to a value of 3, ignoring any user-defined threshold in values.yaml.
2. Missing startupProbe (Other Components): As referenced in Issue Request - Make the startupProbe behavior consistent with liveness and readiness #251, the startupProbe block is missing entirely from the templates for Workflows, Workers, and Jobs.

---

Reproduction Steps

1. Verification of missing failureThreshold in Backend

Define a custom startupProbe with a high failureThreshold in values.yaml:

startupProbe:
  enabled: true
  failureThreshold: 60

Generate the template and inspect the output:

helm template retool retool/retool --version 6.10.3 | grep -A 10 "startupProbe:"

Observed Output:
The rendered YAML contains initialDelaySeconds, timeoutSeconds, etc., but is missing the failureThreshold key.

Template Evidence:
In retool/templates/deployment_backend.yaml, the startupProbe block (approx. line 312) lacks a mapping for .Values.startupProbe.failureThreshold.

2. Verification of missing probes in Workflows

Inspect the Workflows deployment template:

grep "startupProbe" retool/templates/deployment_workflows.yaml
# Result: (empty)

---

Impact

User-defined startup failure thresholds are ignored. Pods are terminated after 3 failures (Kubernetes default), which is often insufficient for Retool environments performing schema migrations. Users are currently forced to increase the livenessProbe.failureThreshold as a workaround, which reduces the responsiveness of health monitoring during normal operation.

---

Proposed Fix

1. Update Backend Deployment (templates/deployment_backend.yaml)

Include the missing key in the startupProbe definition:

        startupProbe:
          ...
          periodSeconds: {{ .Values.startupProbe.periodSeconds }}
          failureThreshold: {{ .Values.startupProbe.failureThreshold }}

2. Standardize Other Deployments

Add the complete startupProbe block to the following templates to ensure consistency:

• templates/deployment_workflows.yaml
• templates/deployment_workers.yaml
• templates/deployment_jobs.yaml
• templates/deployment_code_executor.yaml