Scenario lifecycle: raise, exec, snapshot, restore, destroy
A declaration goes in and a disposable mesh comes out. Verified on a workstation, not asserted: two machines raised and addressed in 14.6s, snapshot 0.28s, restore-to-usable 11.6s, both families pinging with no loss, and the workstation with no route into any of it. The declaration layer implements the model in full — three positions a machine can be in, keyed on forwardability; gateways carrying the address the world sees them as; both address families; multi-homing; MTU; inter-segment policy. It is validated hard because the failures it prevents are silent: a private range on a public segment produces no error, the mesh simply never forms. Public segments are refused unless they use RFC 5737 or RFC 3849 space, and a range wider than the reserved block is refused too. 33 tests, all offline. The runtime implements less than the model, and refuses the difference. A scenario declaring gateways, published ports, policy, inbound deny or place is rejected at raise with every gap named. Raising it would produce a mesh that silently lacks what it declared, which is the fault this lab exists to catch — 04-ISSUES/003, where a firewall key is declared in five manifests and read by no code. Three bugs found by review and by running it, all of one family: The readiness check truthiness-tested incusOk's return. `exec … true` succeeds with EMPTY output, so every machine reported unreachable while incus exec on it worked perfectly. succeeds() now exists so the mistake is not available, and network delete had the same bug — it counted zero segments removed while removing them. list() split instance from machine on the last dash, so a machine called home-server absorbed half the instance id and destroy found nothing. Resources are now found by the metadata they carry, never by name. restore reported success in 0.79s while the machine's agent was still starting, so the next command failed. Both raise and restore now wait for usable and say how long that took — reporting the earlier number is transport reported as effect, which is the fault the lab is being built to find. Two incus behaviours worth recording. Its CLI reads a YAML definition from stdin when stdin is not a terminal, so a spawned command hangs until the timeout kills it and arrives with empty stderr — a failure with no explanation, on a command that works when typed. And it assigns a MAC at runtime without recording it in device config, so MACs are derived and set explicitly, which the guest needs anyway: it names interfaces by bus position, and matching by name configures the wrong one on a multi-homed machine. No build step; Node strips the types. The lifecycle has no unit tests because a fake hypervisor would assert that the fake behaves as expected, which is the shape of test this project exists to stop shipping.
This commit is contained in:
@@ -0,0 +1,267 @@
|
||||
/**
|
||||
* Every rule the design states about a scenario, checked before anything is raised.
|
||||
*
|
||||
* The reason validation is this strict is that the failures it prevents are silent. A
|
||||
* private range on a segment meant to be public does not produce an error — the mesh
|
||||
* simply never forms, because its own code decides public-versus-private by matching the
|
||||
* address. Publishing through a gateway that cannot forward does not produce an error
|
||||
* either; it produces a machine that looks reachable and is not.
|
||||
*
|
||||
* So: refuse the declaration, loudly, before spending a minute raising something that
|
||||
* would have taught us the wrong thing.
|
||||
*
|
||||
* See novox/hq 03-DESIGN/01-to-be/02-scenario-declaration.md
|
||||
*/
|
||||
|
||||
import type { Scenario, Segment } from "./types.ts";
|
||||
import { contains, familyOf, parseAddress, parseCidr, type Cidr } from "./net.ts";
|
||||
|
||||
/** RFC 5737 and RFC 3849. The only addresses guaranteed never to route on the real internet. */
|
||||
const DOCUMENTATION_RANGES = [
|
||||
"192.0.2.0/24",
|
||||
"198.51.100.0/24",
|
||||
"203.0.113.0/24",
|
||||
"2001:db8::/32",
|
||||
].map(parseCidr);
|
||||
|
||||
export class DeclarationError extends Error {
|
||||
readonly problems: string[];
|
||||
|
||||
constructor(problems: string[]) {
|
||||
super(`scenario declaration is not valid:\n - ${problems.join("\n - ")}`);
|
||||
this.name = "DeclarationError";
|
||||
this.problems = problems;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Whether a range sits inside documentation space. Compared as ranges rather than as a
|
||||
* sample address so that a range wider than the reserved block — `203.0.0.0/8`, say — is
|
||||
* correctly refused instead of passing because its first address happens to fall inside.
|
||||
*/
|
||||
function isDocumentationRange(range: Cidr): boolean {
|
||||
return DOCUMENTATION_RANGES.some(
|
||||
(allowed) =>
|
||||
allowed.family === range.family &&
|
||||
range.prefix >= allowed.prefix &&
|
||||
containsRange(allowed, range),
|
||||
);
|
||||
}
|
||||
|
||||
/** Is `inner` entirely within `outer`? Both already parsed, so no address parsing can throw. */
|
||||
function containsRange(outer: Cidr, inner: Cidr): boolean {
|
||||
const bits = outer.family === "v4" ? 32n : 128n;
|
||||
const hostBits = bits - BigInt(outer.prefix);
|
||||
return ((inner.base >> hostBits) << hostBits) === outer.base;
|
||||
}
|
||||
|
||||
/** Ranges of a segment, parsed once, with malformed entries reported rather than thrown. */
|
||||
function rangesOf(name: string, segment: Segment, problems: string[]): Cidr[] {
|
||||
const ranges: Cidr[] = [];
|
||||
for (const text of segment.cidr) {
|
||||
try {
|
||||
ranges.push(parseCidr(text));
|
||||
} catch (err) {
|
||||
problems.push(`segment '${name}': ${(err as Error).message}`);
|
||||
}
|
||||
}
|
||||
return ranges;
|
||||
}
|
||||
|
||||
function inAnyRange(ranges: Cidr[], address: string): boolean {
|
||||
return ranges.some((range) => contains(range, address));
|
||||
}
|
||||
|
||||
export function validate(scenario: Scenario): void {
|
||||
const problems: string[] = [];
|
||||
const segmentNames = Object.keys(scenario.segments);
|
||||
|
||||
if (!scenario.scenario) problems.push("scenario has no name");
|
||||
if (segmentNames.length === 0) problems.push("scenario declares no segments");
|
||||
if (Object.keys(scenario.machines).length === 0) {
|
||||
problems.push("scenario declares no machines");
|
||||
}
|
||||
|
||||
const ranges = new Map<string, Cidr[]>();
|
||||
for (const [name, segment] of Object.entries(scenario.segments)) {
|
||||
ranges.set(name, rangesOf(name, segment, problems));
|
||||
}
|
||||
|
||||
for (const [name, segment] of Object.entries(scenario.segments)) {
|
||||
// A public segment stands in for the internet. Anything that is not documentation
|
||||
// space could route somewhere real, and — far more likely — a private range here
|
||||
// makes the mesh's own public-versus-private test fail silently.
|
||||
if (segment.kind === "public") {
|
||||
// Only ranges that parsed — a malformed one is already reported, and re-parsing it
|
||||
// here would throw out of validation with a single cryptic message instead of the
|
||||
// full list.
|
||||
for (const range of ranges.get(name) ?? []) {
|
||||
if (!isDocumentationRange(range)) {
|
||||
problems.push(
|
||||
`segment '${name}' is public but '${range.text}' is not documentation space ` +
|
||||
`(RFC 5737 / RFC 3849). A non-documentation range here either routes somewhere ` +
|
||||
`real, or — if private — makes the mesh silently never form.`,
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (segment.mtu !== undefined && (segment.mtu < 576 || segment.mtu > 9000)) {
|
||||
problems.push(`segment '${name}': mtu ${segment.mtu} is outside any plausible range`);
|
||||
}
|
||||
|
||||
const gateway = segment.gateway;
|
||||
if (!gateway) continue;
|
||||
|
||||
if (!segmentNames.includes(gateway.to)) {
|
||||
problems.push(`segment '${name}': gateway points at unknown segment '${gateway.to}'`);
|
||||
continue;
|
||||
}
|
||||
if (gateway.to === name) {
|
||||
problems.push(`segment '${name}': gateway points at itself`);
|
||||
continue;
|
||||
}
|
||||
|
||||
// The gateway's address is what the world sees this network as, so it belongs to the
|
||||
// PARENT segment, not this one. Getting this backwards is easy and produces a topology
|
||||
// that raises fine and reproduces nothing.
|
||||
const parentRanges = ranges.get(gateway.to) ?? [];
|
||||
for (const address of gateway.address) {
|
||||
try {
|
||||
parseAddress(address);
|
||||
} catch (err) {
|
||||
problems.push(`segment '${name}' gateway: ${(err as Error).message}`);
|
||||
continue;
|
||||
}
|
||||
if (parentRanges.length > 0 && !inAnyRange(parentRanges, address)) {
|
||||
problems.push(
|
||||
`segment '${name}' gateway: address '${address}' is not within '${gateway.to}' ` +
|
||||
`(${scenario.segments[gateway.to]?.cidr.join(", ")}). A gateway's address is the ` +
|
||||
`one the parent network sees, not one on the segment behind it.`,
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
for (const family of gateway.nat) {
|
||||
if (family !== "v4" && family !== "v6") {
|
||||
problems.push(`segment '${name}' gateway: '${family}' is not an address family`);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// A gateway chain must terminate. A cycle would raise forever rather than fail.
|
||||
for (const name of segmentNames) {
|
||||
const seen = new Set<string>([name]);
|
||||
let current = scenario.segments[name]?.gateway?.to;
|
||||
while (current) {
|
||||
if (seen.has(current)) {
|
||||
problems.push(`segment '${name}': gateway chain loops through '${current}'`);
|
||||
break;
|
||||
}
|
||||
seen.add(current);
|
||||
current = scenario.segments[current]?.gateway?.to;
|
||||
}
|
||||
}
|
||||
|
||||
for (const [name, machine] of Object.entries(scenario.machines)) {
|
||||
if (machine.at === "detached") {
|
||||
if (machine.published?.length) {
|
||||
problems.push(`machine '${name}' is detached but declares published ports`);
|
||||
}
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!Array.isArray(machine.at) || machine.at.length === 0) {
|
||||
problems.push(`machine '${name}': 'at' must be an attachment, a list of them, or "detached"`);
|
||||
continue;
|
||||
}
|
||||
|
||||
const attachedTo = new Set<string>();
|
||||
for (const attachment of machine.at) {
|
||||
if (!segmentNames.includes(attachment.segment)) {
|
||||
problems.push(`machine '${name}': unknown segment '${attachment.segment}'`);
|
||||
continue;
|
||||
}
|
||||
if (attachedTo.has(attachment.segment)) {
|
||||
problems.push(`machine '${name}': attached to '${attachment.segment}' more than once`);
|
||||
}
|
||||
attachedTo.add(attachment.segment);
|
||||
|
||||
const segmentRanges = ranges.get(attachment.segment) ?? [];
|
||||
const seenFamilies = new Set<string>();
|
||||
for (const address of attachment.address) {
|
||||
try {
|
||||
parseAddress(address);
|
||||
} catch (err) {
|
||||
problems.push(`machine '${name}': ${(err as Error).message}`);
|
||||
continue;
|
||||
}
|
||||
const family = familyOf(address);
|
||||
if (seenFamilies.has(family)) {
|
||||
problems.push(`machine '${name}': two ${family} addresses on '${attachment.segment}'`);
|
||||
}
|
||||
seenFamilies.add(family);
|
||||
|
||||
if (segmentRanges.length > 0 && !inAnyRange(segmentRanges, address)) {
|
||||
problems.push(
|
||||
`machine '${name}': address '${address}' is not within segment ` +
|
||||
`'${attachment.segment}' (${scenario.segments[attachment.segment]?.cidr.join(", ")})`,
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
for (const publication of machine.published ?? []) {
|
||||
if (!Number.isInteger(publication.port) || publication.port < 1 || publication.port > 65535) {
|
||||
problems.push(`machine '${name}': port ${publication.port} is not a port`);
|
||||
}
|
||||
const via = scenario.segments[publication.on];
|
||||
if (!via) {
|
||||
problems.push(`machine '${name}': publishes on unknown segment '${publication.on}'`);
|
||||
continue;
|
||||
}
|
||||
if (!attachedTo.has(publication.on)) {
|
||||
problems.push(
|
||||
`machine '${name}': publishes on '${publication.on}' but is not attached to it`,
|
||||
);
|
||||
continue;
|
||||
}
|
||||
if (!via.gateway) {
|
||||
problems.push(
|
||||
`machine '${name}': publishes on '${publication.on}', which has no gateway to ` +
|
||||
`forward through`,
|
||||
);
|
||||
continue;
|
||||
}
|
||||
// The constraint being reproduced, not an implementation limit: a machine behind a
|
||||
// gateway it does not control cannot be published, and pretending otherwise would
|
||||
// make the lab certify something production cannot do.
|
||||
if (!via.gateway.forwardable) {
|
||||
problems.push(
|
||||
`machine '${name}': cannot publish through '${publication.on}' — its gateway is ` +
|
||||
`not forwardable. That is the constraint being reproduced, not a limitation.`,
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
for (const rule of scenario.policy ?? []) {
|
||||
for (const side of [rule.from, rule.to]) {
|
||||
if (!segmentNames.includes(side)) {
|
||||
problems.push(`policy: unknown segment '${side}'`);
|
||||
}
|
||||
}
|
||||
if (rule.from === rule.to) {
|
||||
problems.push(`policy: '${rule.from}' to itself is not a rule`);
|
||||
}
|
||||
}
|
||||
|
||||
for (const machine of Object.keys(scenario.place ?? {})) {
|
||||
if (machine === "all") continue;
|
||||
if (!(machine in scenario.machines)) {
|
||||
problems.push(`place: '${machine}' is not a machine in this scenario`);
|
||||
}
|
||||
}
|
||||
|
||||
if (problems.length > 0) throw new DeclarationError(problems);
|
||||
}
|
||||
Reference in New Issue
Block a user