On March 17, 2026, we experienced a public networking incident affecting Virtual Machines in India (BOM2). The issue was caused by a downstream provider routing problem, which led to a portion of public traffic being misrouted unexpectedly. As a result, some customer instances experienced failed public connectivity, intermittent packet loss, unstable network behavior, and in some cases reduced network throughput or degraded speeds.

The impact varied depending on traffic pattern and workload. For some users, services became partially unreachable over the public internet. Others may have seen slower response times, inconsistent connectivity, or disrupted access to applications exposed through public networking. Internal VM operation was not broadly affected, but traffic relying on public ingress or egress in the affected path could experience degradation.

We first identified the incident as a public routing problem in BOM2 and began active mitigation immediately. During investigation, we confirmed that the issue was not caused by guest virtual machines themselves, but by an upstream routing fault in the downstream provider path serving public networking. Once the faulty route behavior was isolated and corrected, network paths stabilized and connectivity returned to normal.

The issue is now fully resolved. Public networking in BOM2 has been restored, and affected Virtual Machine services should now be operating normally.

We understand incidents like this directly affect reliability and user trust. We are reviewing the event with the provider and internally to improve detection, escalation, and path-level resilience so similar routing failures can be identified and mitigated faster in the future.