Between May 12 and May 14, 2026, Akamai Cloud Pulse (ACLP) experienced two related service interruptions that temporarily disrupted the ingestion and display of telemetry metrics specifically for Managed Databases.

• Window 1: May 12, 23:15 UTC – May 13, 00:11 UTC
• Window 2: May 14, 06:47 UTC – 07:33 UTC

During these windows, customers were unable to view Managed Database performance metrics within the Akamai Cloud Manager dashboards or retrieve them via external API and Collector streams. Metric-based alerting for Managed Databases was also impaired during both impact windows. We are currently evaluating whether the metrics data missing from these windows can be successfully backfilled.

Remark: Metrics for other services, as well as the Log function, were unaffected and operated normally. Furthermore, this incident was strictly limited to the observability reporting layer. The performance, availability, and data integrity of customers’ underlying Managed Databases were never impacted.

The disruption was caused by a cascading failure initiated by a brief network routing issue.

During the initial event, a localized network disruption caused a small percentage of metrics traffic to fail to reach our processing endpoints. While the network disruption was brief, the accumulated backlog of automatic retries from the telemetry sender's built-in retry mechanism created a sudden, massive spike in traffic against the ingest endpoint. 

This "retry storm" overloaded the specific downstream message-queuing infrastructure responsible for processing and storing Managed Database telemetry. The processing pipeline was unable to handle the surge, causing the components to fail and leading to the complete loss of DBaaS ingestion (Window 1). The degraded performance observed two days later (Window 2) was the result of a separate, independent infrastructure incident within our telemetry partner's message-queuing service, entirely unrelated to the initial network disruption. 

To mitigate the issue, our engineering teams, working alongside our telemetry infrastructure partner, intervened to scale up the processing pipeline's configuration to absorb and clear the accumulated retry backlog. Following these actions, Managed Database metrics ingestion returned to normal levels.

This summary reflects our current understanding of the incident based on available information. Our investigation is ongoing, and details may change as we learn more.