Tailscale v1.102.3
Update instructionsAll Platforms
- Go is updated to version 1.26.6.
- Tailscale refuses host-scoped IPv4 destinations at every point that acts on an unmapped 4via6 address. This fix addresses a security vulnerability described in TS-2026-011.
- When MagicDNS is disabled, unqualified hostnames are correctly forwarded to the configured nameservers.
- Resolved an issue where nodes with Tailnet Lock enabled on large tailnets would experience startup failures.
Windows
- Resolved an issue where the Windows installer would fail during an upgrade when the name of the user's home directory contained a space.
tailscaledreturns an error instead of panicking on devices where IPv6 or NetBios over TCP/IP is disabled.
iOS
- Memory usage is reduced for iOS devices on large tailnets.
tvOS
- Memory usage is reduced for tvOS devices on large tailnets.
Tailscale container image v1.102.3
A new release of the Tailscale container image is available. You can download it from Docker Hub or from our GitHub packages repository.
- The container waits longer for its initial connection to the coordination server when you set the new
TS_BOOT_TIMEOUTenvironment variable. It accepts a duration, such as90sor3m, and defaults to60s.
Tailscale Kubernetes Operator v1.102.3
A new release of the Tailscale Kubernetes Operator is available. For guidance on installing and updating, refer to our installation instructions.
- Peer relays on AWS are reachable regardless of the availability zone their pod is scheduled in. Pinning Elastic IPs also requires pinning the pods to the same zone using a ProxyClass.
- Network Load Balancers fronting peer relays on AWS use HTTP instead of TCP for pod health checks. Peer relays listen only on UDP, so the default TCP health check could never succeed, causing every target to be reported as unhealthy.