mesh-lab/replays/register.go

// Package replays is the lab's replay of every core incident (novox/hq to-be 45 §9, M9): a scripted
// replay of what happened, asserting the rule's outcome, run before every merge of a core repository —
// and proved, once, to fail on the commit before its fix and to pass on the fix.
//
// **Where a replay lives.** A replay of one component's logic is a test in that component's repository,
// written only with what the component had before its fix, so it can be laid over the older commit
// (the controller's replays_test.go). A replay of what the mesh runs rather than what it wrote — the bus
// server's release, the resolver the catalogue configures, under both C libraries — lives here, where a
// container runtime and a bus of a given release are at hand. Every one is in Register, with the issue,
// the fix, and how the prover runs it at a commit.
//
// **A new core issue resolves with its replay added here, or with a stated reason none is possible**
// (hq 00-META/checks/cycle.py holds the issue to it: `replay:` names an entry here, or `replay-none:`
// says why).
package replays

// Kind is how a replay is run at a commit.
type Kind string

const (
	// InRepository is a test in the fixed repository, laid over the commit and run there with go test.
	InRepository Kind = "test"
	// Bus is the bus replay here, run against a server of the release the catalogue pinned at the commit.
	Bus Kind = "bus"
	// Resolver is the resolver replay here, run with the catalogue's resolver configuration at the commit,
	// asked by a program under musl and one under glibc.
	Resolver Kind = "resolver"
	// Gate is a test of the merge gate itself: before its fix there was no check to fail, so the commit
	// before is held to fail by not having it.
	Gate Kind = "gate"
	// Liveness is the crash-loop replay here: a container that exits at start, raised on the runtime here,
	// judged by the node-engine at the commit, and the controller at the commit judging its gate from what
	// the engine said (novox/hq ADR 0240). Two repositories move together: Repository and With.
	Liveness Kind = "liveness"
)

// Also is another repository a replay runs at its own commits beside Repository's: before its fix and on
// it, as Repository is.
type Also struct {
	Repository string
	Fix        string
	// Before is the commit the replay runs this repository at before the fix, when that is not Fix's
	// first parent.
	Before string
}

// Replay is one incident, replayed.
type Replay struct {
	ID    string
	Issue int
	// What is the incident in a line, and Asserts the rule's outcome the replay holds.
	What, Asserts string
	Kind          Kind
	// Repository is where the fix is, and Fix the fix's commit there: the replay must fail on Fix's first
	// parent and pass on Fix.
	Repository string
	Fix        string
	// Before is the commit the replay must fail on, when that is not Fix's first parent: a fix still on its
	// branch is proved against the main it branched from. Set to "" once the fix is merged.
	Before string
	// Package and Test are where the replay is run: a package of Repository for InRepository and Gate,
	// of this module otherwise; Files are the files laid over the commit, from where the replay lives.
	Package string
	Test    string
	Files   []string
	// Home is where the replay's files live, when that is not the fix: the repository and ref.
	Home, HomeRef string
	// With is the other repositories the replay runs at their commits beside Repository's (Liveness).
	With []Also
}

// Register is every replay, by incident.
var Register = []Replay{
	{ID: "R236", Issue: 236, Kind: Gate, Repository: "mesh-controller", Fix: "feat/merge-gate",
		Before:  "origin/main",
		What:    "a login manager's service passed the catalogue check and was refused whole by the node-engine",
		Asserts: "a manifest the node-engine refuses fails its pull request, naming the machine and the module",
		Package: "./cmd/mesh-controller", Test: "TestIssue236", Files: []string{"cmd/mesh-controller/merge_gate_test.go"}},
	{ID: "R262", Issue: 262, Kind: Resolver, Repository: "mesh-catalog", Fix: "20603b63e67323f2c725b242c0ea87281d4dee93",
		What:    "an Alpine container could not find a machine by its mesh name: NXDOMAIN for IPv6 is final to musl",
		Asserts: "every machine's name resolves through the catalogue's resolver under musl and under glibc",
		Package: ".", Test: "TestReplay262"},
	{ID: "R263", Issue: 263, Kind: InRepository, Repository: "mesh-controller", Fix: "6d620f77c3f3f967be953ff760823c6a51b7b8d3",
		What:    "a consumer's 26-character identity refused the resolver holder's whole declaration",
		Asserts: "a provider's machine composes whatever its consumers are called",
		Package: "./cmd/mesh-controller", Test: "TestReplay263", Files: []string{"cmd/mesh-controller/replays_test.go"},
		Home: "mesh-controller", HomeRef: "feat/replays"},
	{ID: "R266", Issue: 266, Kind: Bus, Repository: "mesh-catalog", Fix: "0275c2e",
		What:    "a merge on the bus was never handed to the controller: 2.10 skipped messages on a consumer with several filters",
		Asserts: "a consumer filtered like the controller's is handed every followed message under the mesh's traffic",
		Package: ".", Test: "TestReplay266"},
	{ID: "R273", Issue: 273, Kind: InRepository, Repository: "mesh-controller", Fix: "8bfaf1523ed2dddde80daa0994f1044e59db44b1",
		What:    "a rule for the resolver re-bound a machine's database consumers to empty databases elsewhere",
		Asserts: "a consumer beside its store stays bound to it while another machine holds the store's seat",
		Package: "./cmd/mesh-controller", Test: "TestReplay273", Files: []string{"cmd/mesh-controller/replays_test.go"},
		Home: "mesh-controller", HomeRef: "feat/replays"},
	// Not a core incident, and the first of its kind: a module's. The crash loop was found by a person
	// reading the agent server's log for another reason (issue 268); research 032 measured it, and ADR
	// 0240 makes the node-engine judge it and the gate fail it.
	{ID: "R-crashloop", Issue: 268, Kind: Liveness, Repository: "mesh-controller",
		Fix: "feat/a-module-says-how-it-is-healthy", Before: "origin/main",
		With: []Also{{Repository: "mesh-host", Fix: "feat/a-module-says-how-it-is-healthy", Before: "origin/main"}},
		What: "the agent server crash-looped about a hundred times while the mesh read it applied, its tools served " +
			"and nothing raised: the gate judged a module by what the mesh saw from outside",
		Asserts: "a container whose program exits at start is said unhealthy by the node-engine, and its build fails " +
			"its gate on the first machine and is put back there",
		Package: ".", Test: "TestReplayCrashLoop", Files: []string{"cmd/mesh-controller/replays_test.go"}},
	// A module's again: the silent web application, which no liveness and no TCP check sees, raised by the
	// HTTP check its module declares (ADR 0240 Phase B).
	{ID: "R145", Issue: 145, Kind: Liveness, Repository: "mesh-controller",
		Fix: "feat/health-the-field", Before: "origin/main",
		With: []Also{{Repository: "mesh-host", Fix: "feat/health-the-field", Before: "origin/main"}},
		What: "a web application accepted TCP and answered nothing for eleven hours while the mesh said its machine " +
			"healthy: its port was open and its program ran",
		Asserts: "a web server whose application never answers is said unhealthy by its declared HTTP check within two " +
			"looks, and the controller raises the module's condition on the second",
		Package: ".", Test: "TestReplaySilentWebApp", Files: []string{"cmd/mesh-controller/replays_test.go"}},
}

// Find is the replay of that id, or of that issue.
func Find(id string) (Replay, bool) {
	for _, r := range Register {
		if r.ID == id || r.ID == "R"+id {
			return r, true
		}
	}
	return Replay{}, false
}