During the incident, multiple API proxies returned HTTP 502 responses for approximately 5 minutes. Google Cloud HTTP Load Balancer logs consistently show statusDetails="failed_to_connect_to_backend"
The failed requests do not appear in Apigee API Monitoring and never reached our backend applications (no application logs were generated). The issue resolved without any configuration changes.
We would like Google Support to investigate whether there was a transient issue affecting the managed Apigee X infrastructure or connectivity between the Google-managed HTTP(S) Load Balancer and the Apigee runtime in the asia-southeast1 region during the incident window.
Hi @sfathima, welcome back to the Apigee community, it’s great to see you again!
We have seen your question regarding the intermittent 502 errors. Since this may require further investigation, we highly recommend opening a ticket with Google Cloud Support so they can look into the details of the issue. However, we are leaving this thread open and encourage the community to share if they have experienced a similar issue.
In the meantime, we’d love to invite you to our upcoming Community Tech Talk tomorrow, Thursday, August 13th at 12:00 p.m. NYC time. We will be discussing Secure Agent MCP Tool Governance using Apigee. Our speaker, Tarun Abraham, will be giving a real-time walkthrough of payload parsing using the new ParsePayload policy and showing how to eliminate security blind spots in Agentic Tool workflows.
Hi @sfathima, we’ve seen a similar pattern on Apigee X.
When 502s don’t appear in Apigee API Monitoring at all, it’s a strong signal the failure is occurring between the GCP HTTP(S) Load Balancer and the Apigee runtime — not in your proxy logic or backend services.
A few things worth checking while the support ticket is being investigated:
Filter your LB access logs on statusDetails="failed_to_connect_to_backend" to confirm the fault layer
Cross-reference the GCP Status Dashboard for your region (asia-southeast1) during the exact incident window
Set up a separate alert on LB-level 5xx rate — this class of failure is invisible to Apigee API Monitoring, so you need LB metrics as a parallel signal
For the support ticket, including the LB backend service name alongside the time window and region will help Google pinpoint whether it was a transient issue at the Apigee runtime MIG level.
