[RELEASE] ScyllaDB 2026.1.8

Release Notes
, ,
1

The ScyllaDB team is pleased to announce the release of ScyllaDB 2026.1.8, a production-ready patch release for ScyllaDB 2026.1 Long Term Support (LTS) Release.

More information on ScyllaDB Version Support Policy is available here.

Related Links

Bug Fixes

The following issues are fixed in this release.

Alternator

  • Alternator TTL expiration scans on tablet-backed tables now release the effective replication map (ERM) between individual tablet scans. Previously, a long-running TTL scan could hold a reference to stale topology metadata for the duration of the entire table scan, which could prevent old token metadata versions from being released and delay tablet rebalancing operations waiting in barrier_and_drain. Each tablet’s scan now re-resolves the current ERM and copies only the needed partition range before releasing the metadata reference, ensuring that topology changes proceed without unnecessary delays.
    scylladb#30540

Networking

  • Cached RPC connections are now invalidated when a peer node’s IP address changes. Previously, when a node restarted with a new IP address (e.g. during Kubernetes pod rescheduling), gossip updated the address map correctly but the messaging service continued using cached connections pointing to the old IP. Because the old IP is typically blackholed in container environments, the stale connection would not fail immediately, causing the direct failure detector to mark the restarted node as down until the TCP keepalive timeout expired. The CLIENT_ID handler now detects IP changes and evicts stale connections on all shards before updating the address map, so subsequent RPCs are routed to the correct address immediately.
    scylladb#30620

Raft

  • A race condition during concurrent node joins is now handled correctly. When many nodes join a cluster simultaneously and some joining nodes are configured as seeds for other joining nodes, a node could initiate a group0 operation immediately after starting the group0 server but before updating its upgrade state to use_post_raft_procedures. This resulted in a legacy guard being returned with zero semaphore units, which triggered an assertion failure in add_entry. The start_operation logic now properly handles this transitional state.
    scylladb#30522

  • The view building coordinator no longer uses the Raft timeout when starting a group0 operation for finished-task garbage collection. Because the coordinator already runs on the Raft leader, a timeout is unnecessary, and on heavily loaded nodes the read barrier could take longer than the default Raft timeout, causing the garbage collection fiber to time out and log errors during view drop operations. Removing the timeout allows the operation to complete regardless of momentary load spikes.
    scylladb#30610

Service Levels

  • The computation of effective service levels now gracefully handles cycles in the role grants graph. Under normal operation the role grants graph is acyclic, but inconsistent state migrated from pre-Raft authorization or manual edits could introduce cycles that caused the previous topological sort algorithm to fail with an exception, breaking service level updates on the affected node. The algorithm is replaced with a depth-first search with memoization that detects and reports cycles via a warning log while still computing a usable set of effective service level parameters.
    scylladb#30399