Last week in scylla-cluster-tests.git master (issue #143; 2026-07-03)

This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the fa58b5b1…6855a9fa range are covered.

There were 35 non-merge commits from 12 authors in that period. Some notable commits:

Added AWS region fallback on capacity exhaustion: when a region runs out of capacity during provisioning, SCT now relocates the whole cluster to the next eligible region instead of failing. The multi-DC variant relocates only the exhausted DC while healthy DCs stay in place, and the feature was subsequently enabled by default.

A new implementation plan for feature-aware adaptive topology timeouts was registered, proposing schema-based detection of MV, SI, CDC, and LWT so decommission, new-node, and tablet-migration operations get larger (up to 3x) time budgets when those features are enabled, with Argus telemetry for post-failure diagnosis.

The scylla-doctor bump to v1.12 (from 1.10) brings several new collectors and analyzers — large-partitions/cells/rows and role-permissions collectors, STCS-in-schema and keyspace-replication analyzers — plus aarch64 support across the architecture, clock-source, and instruction-set analyzers and coverage for newer AWS instance types (i7i/i7ie/i8g/i8ge).

On the infrastructure side, the SCT runner base image was bumped to Ubuntu 26.04. The cloud catalog also dropped Azure Dpdsv5 from pricing/sizing since it has no local NVMe disk that scylla-machine-image can use.

Backport automation migrated to the shared github-automation reusable workflow used by scylla-pkg, scylla-dtest, and scylla-machine-image — backports now target branch-X.Y directly and run in parallel by default, with a later pin bump enabling a cascade_backport opt-out label for chained backports.

Argus integration gained an always-on request replay log: every mutating Argus API call is now journaled to JSONL so it can be replayed later in case of Argus downtime or issues. Brought also new hydra investigate download-logs subcommand.

See you in the next issue of last week in scylla-cluster-tests.git master!