LaunchDarkly is the leading feature flag and feature management platform, used by engineering teams at companies like IBM, Atlassian, and Intuit. When LaunchDarkly's streaming service has issues, it raises a critical question: are your feature flags still working, or has your kill switch just become unreachable?
The Good News: LaunchDarkly is Built for Resilience
Unlike most SaaS dependencies, a LaunchDarkly outage typically doesn't break your application. Here's why: LaunchDarkly SDKs (server-side and client-side) maintain a local in-memory cache of all flag values. If the streaming connection to LaunchDarkly's servers drops, your SDK continues evaluating flags using the last known state.
✅ What Still Works During an Outage
- Flag evaluations (using cached values)
- Your application serving users normally
- Any flag values set before the outage
❌ What Breaks During an Outage
- Pushing new flag changes to production
- Activating kill switches or emergency rollbacks
- Changing targeting rules or percentage rollouts
- Accessing the LaunchDarkly dashboard
How to Diagnose a LaunchDarkly Outage
1. Check the Official Status Page
Visit status.launchdarkly.com for the authoritative view. LaunchDarkly breaks down status by component: Streaming API, Evaluation API (relay proxy), Dashboard, and Experiments. A streaming outage is the most impactful for real-time flag updates.
2. Check Your SDK Logs
LaunchDarkly SDKs log connection events. Look for messages like "Encountered a network error" or "Streaming disabled" in your application logs. These indicate the SDK has lost its streaming connection and is operating in cached mode.
📡 Monitor LaunchDarkly uptime every 30 seconds — get alerted in under a minute
Trusted by 100,000+ websites · Free tier available
3. Test the Streaming Endpoint
You can manually verify the LaunchDarkly streaming endpoint:
curl -s -o /dev/null -w "%{http_code}" \
https://stream.launchdarkly.com/all \
-H "Authorization: YOUR_SDK_KEY"
# Expect 200 or 401 if key is invalid (not 5xx)Monitor your LaunchDarkly integration
Set up uptime checks on LaunchDarkly's streaming endpoint so you're immediately alerted when your team loses the ability to push flag changes.
Try Better Stack Free →Why LaunchDarkly Goes Down: Common Causes
- Streaming Service Overload: High-volume event ingestion (experiment data, click tracking) can cause streaming latency or disconnections.
- Database Failovers: Flag storage database maintenance or failover events can cause brief flag delivery interruptions.
- AWS Dependency: LaunchDarkly runs on AWS. Regional outages can affect flag delivery latency.
- DDoS or Traffic Spike: The Evaluation API endpoint (used by SDKs with relay proxy) can be affected by unusual traffic patterns.
Using the LaunchDarkly Relay Proxy for Resilience
The LaunchDarkly Relay Proxy is a self-hosted component that sits between your SDK and LaunchDarkly's servers. If you run the relay proxy, your application can continue receiving flag updates from your own infrastructure even if LaunchDarkly's cloud service is degraded.
Enterprise teams with strict uptime requirements should strongly consider running the relay proxy. It adds a layer of resilience and can reduce latency for high-traffic applications.
Frequently Asked Questions
Is LaunchDarkly down right now?
Check status.launchdarkly.com for real-time status. If you see a streaming disruption, your application flags are still working from cache — but you cannot push new flag changes until service is restored.
How do I set default flag values for when LaunchDarkly is unreachable?
All LaunchDarkly SDK variation calls accept a fallback value parameter. This is the value returned when the SDK cannot evaluate the flag (e.g., during initialization before the first data load). Always set meaningful defaults.
Should I use LaunchDarkly as a kill switch during incidents?
Yes — but be aware that kill switches only work if LaunchDarkly itself is available. If LaunchDarkly is down during your incident, you cannot use it for emergency rollbacks. Plan for this with manual feature flag overrides in your codebase as a backstop.
Alert Pro
14-day free trialStop checking — get alerted instantly
Next time LaunchDarkly goes down, you'll know in under 60 seconds — not when your users start complaining.
- Email alerts for LaunchDarkly + 9 more APIs
- $0 due today for trial
- Cancel anytime — $9/mo after trial