This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the 868b3ff0…b2093b44 range are covered.
There were 30 non-merge commits from 11 authors in that period. Some notable commits:
Cloud provisioning saw a wave of capacity-awareness work across backends. AWS gained automatic AZ fallback on capacity errors in the modern provisioning path, gated by fallback_to_next_availability_zone , default true. GCE got the same treatment with upfront AZ filtering and dynamic zone discovery.
The AWS VPC peering layer was rewritten with VPC-ID-based lookup, proper accept_vpc_peering_connection() with retries on the post-creation timing window, CreatedBy=SCT tagging, programmatic cleanup of blackhole routes and rejected peerings, and a dry_run mode. Fixes the long-standing root cause of unaccepted peerings (target VPC’s accept() swallowing OperationNotPermitted).
The Migrator integration now reports P99 write latency, throughput, dataset size, speed and duration to Argus. xcloud’s weekly provisioning trigger gained VS-on-ScyllaCloud node deployment to catch deployment regressions earlier.
The bundled latte version was bumped twice — first to v0.47.0 bringing a rune 0.13 →0.14 upgrade, Alternator API enhancements, a workloads/alternator/type_validation.rn fix and an openssl security update, then to v0.47.1 which switches rune to a fork carrying a compiler-divergence fix and bumps openssl again.
The cql-stress version was updated to v0.2.9, focused on dependencies upgrade.
See you in the next issue of last week in scylla-cluster-tests.git master!