What happened

On February 20, 2026, newly provisioned VMs assigned to subnet 148.113.14.192/28 in the BOM2 (Mumbai) region became unreachable via their public floating IPs. Existing VMs and all other subnets in the region continued to operate normally throughout the incident.

Root cause

During routine infrastructure operations, a configuration inconsistency was introduced on the affected subnet that caused a mismatch between the network layer and our floating IP allocation pipeline. This resulted in floating IPs assigned from this subnet failing to route traffic correctly to the associated VMs.

Resolution

After identifying the affected subnet, our team deprovisioned it from the active networking pool and migrated impacted floating IP assignments to a healthy subnet. New VM provisioning was restored once the subnet was fully removed from rotation. The deprovisioned subnet will not be reintroduced until it has been fully validated.

What we're doing to prevent this

We are adding automated pre-flight validation checks to our subnet provisioning pipeline to catch configuration inconsistencies before they reach production. We are also expanding our network-layer monitoring to detect routing mismatches earlier.