This short report brings to light some interesting commits to [scylla-cluster-tests.git master](GitHub - scylladb/scylla-cluster-tests: Tests for Scylla Clusters · GitHub) from the last week.
Commits in the 3f7410ea…f89678e0 range are covered.
There were 116 non-merge commits from 14 authors in that period. Some notable commits:
SCT can now [estimate what a run will cost before it provisions anything](feature(cost): estimate what a run costs before it provisions anything · scylladb/scylla-cluster-tests@c0a093c · GitHub) with `sct.py estimate-cost`, priced at the lifecycle the run will use. Each run [sends its estimate to Argus](feature(cost): attach the run's cost estimate to Argus · scylladb/scylla-cluster-tests@3a37c46 · GitHub), and every node [reports its actual cost to Argus](feature(cost): report each instance's actual cost to Argus · scylladb/scylla-cluster-tests@1a2e9e0 · GitHub) when it is terminated.
The vendored Argus client was [updated to 0.16.5](chore(argus): update vendored argus client to 0.16.5 · scylladb/scylla-cluster-tests@640f590 · GitHub). The release adds the per-run cost API and the `–show-hidden` and `–event-id` CLI flags, and scopes tunnel keys per run.
A new [`FixedOrderMonkey` runner executes disruptions in the exact order](feat(nemesis): add FixedOrderMonkey runner for reproducible disruptio… · scylladb/scylla-cluster-tests@dfad092 · GitHub) given in its `nemesis_selector` and repeats the sequence once exhausted.
The Docker backend gained [Scylla Manager support](feature(docker): add Scylla Manager support for Docker backend · scylladb/scylla-cluster-tests@4024ef8 · GitHub), so `use_mgmt` with backup and repair operations now works in local Docker runs. Log collection also [gathers `scylla-manager-agent.log`](feature(logcollector): collect scylla-manager-agent.log · scylladb/scylla-cluster-tests@0192f4e · GitHub) from DB nodes.
Tests can now [capture screenshots of specific Grafana panels](feature(logcollector): add configurable Grafana panel screenshot capture · scylladb/scylla-cluster-tests@7b84af8 · GitHub) at log collection, configured per test through the `grafana_screenshot_panels` YAML option.
The trigger matrix now has a weekly [rolling upgrade job for OCI images](feature(rolling-upgrade): add OCI image rolling upgrade job · scylladb/scylla-cluster-tests@fa0cc0d · GitHub), documented in the new `docs/rolling-upgrade.md` guide.
Branch images can now be [pinned by Scylla SHA on all backends](feature(images): pin branch images by Scylla SHA on all backends · scylladb/scylla-cluster-tests@470ce74 · GitHub) with the `:` form, for example `master:bedcc69`. `list-images` shows the Scylla SHA of each image.
Minicloud runs now [lock guest RAM so the host cannot swap it out](feature(minicloud): lock guest RAM so the host cannot swap it out · scylladb/scylla-cluster-tests@04fd67a · GitHub), controlled by the new `minicloud_lock_guest_memory` option. Placement groups and capacity reservations are also [available on minicloud](feature(minicloud): allow placement groups and capacity reservations · scylladb/scylla-cluster-tests@df4a679 · GitHub), and minicloud [rejects instance types of a foreign architecture](feature(minicloud): start minicloud with --cross-arch=fail · scylladb/scylla-cluster-tests@e4efbd8 · GitHub).
Scylla Doctor was [bumped to 1.14.1](chore(deps): update dependency scylla-doctor to v1.14.1 · scylladb/scylla-cluster-tests@28974e1 · GitHub), which keeps the output of output-only collectors in vitals and stores `DESC SCHEMA` in the collected data.
See you in the next issue of last week in scylla-cluster-tests.git master!