import * as kafka from "@distilled.cloud/gcp/managedkafka_v1";
import { createPhysicalName } from "alchemy/PhysicalName";
import { tagRecord } from "alchemy/Tags";
import * as Data from "effect/Data";
import * as Effect from "effect/Effect";
import * as Schedule from "effect/Schedule";
import * as Stream from "effect/Stream";
import { isTransientGcpError } from "../Errors.ts";
import { alchemyLabelKeys, createInternalLabels, stripInternalLabels } from "../Labels.ts";
import { waitForOperation as waitForGcpOperation } from "../Operation.ts";
export const MAX_NAME_LENGTH = 63;
/** The API rejects longer cluster and Connect cluster ids. */
export const MAX_CLUSTER_ID_LENGTH = 53;
export const DEFAULT_VCPU_COUNT = 3;
export const DEFAULT_MEMORY_BYTES = 3_221_225_472;
export const OWNERSHIP_SUBJECT = "alchemy_ownership";
export class ResourceNotResolved extends Data.TaggedError("GCP.ManagedKafka.ResourceNotResolved")<{
resourceName: string;
}> {}
export class ResourceStillExists extends Data.TaggedError("GCP.ManagedKafka.ResourceStillExists")<{
resourceName: string;
}> {}
export class ResourceNotReady extends Data.TaggedError("GCP.ManagedKafka.ResourceNotReady")<{
resourceName: string;
state: string;
}> {}
export class ResourceFailed extends Data.TaggedError("GCP.ManagedKafka.ResourceFailed")<{
resourceName: string;
state: string;
}> {}
export const lastSegment = (value: string) => {
const trimmed = value.replace(/\/+$/, "");
const parts = trimmed.split("/");
return parts[parts.length - 1] || trimmed;
};
export const rfc1035 = (name: string, fallback = "kafka", maxLength = MAX_NAME_LENGTH): string => {
let next = name
.toLowerCase()
.replace(/[^a-z0-9-]/g, "-")
.replace(/-+/g, "-")
.replace(/^-+|-+$/g, "");
if (!/^[a-z]/.test(next)) next = `k${next}`;
next = next.slice(0, maxLength).replace(/-+$/g, "");
if (next.length === 0) return fallback;
if (!/[a-z0-9]$/.test(next)) next = `${next.slice(0, maxLength - 1)}0`;
return next.slice(0, maxLength);
};
export const schemaRegistryIdOf = (name: string, fallback = "schema"): string => {
let next = name
.toLowerCase()
.replace(/[^a-z0-9_]/g, "_")
.replace(/_+/g, "_")
.replace(/^_|_$/g, "");
if (!/^[a-z]/.test(next)) next = `s${next}`;
next = next.slice(0, MAX_NAME_LENGTH).replace(/_+$/g, "");
return next.length > 0 ? next : fallback;
};
export const normalizeLocation = (location: string | undefined, defaultLocation: string) =>
lastSegment(location ?? defaultLocation).toLowerCase();
export const locationParent = (project: string, location: string) =>
`projects/${project}/locations/${location}`;
export const defaultSubnet = (project: string, location: string) =>
`projects/${project}/regions/${location}/subnetworks/default`;
export const toPhysicalId = (
id: string,
explicit: string | undefined,
existing: string | undefined,
fallback = "kafka",
maxLength = MAX_NAME_LENGTH,
) =>
Effect.gen(function* () {
if (explicit !== undefined) return rfc1035(explicit, fallback, maxLength);
if (existing !== undefined) return existing;
return rfc1035(
yield* createPhysicalName({
id,
maxLength,
lowercase: true,
}),
fallback,
maxLength,
);
});
export const toSchemaRegistryId = (
id: string,
explicit: string | undefined,
existing: string | undefined,
) =>
Effect.gen(function* () {
if (explicit !== undefined) return schemaRegistryIdOf(explicit);
if (existing !== undefined) return existing;
return schemaRegistryIdOf(
yield* createPhysicalName({
id,
maxLength: MAX_NAME_LENGTH,
lowercase: true,
delimiter: "_",
}),
);
});
export const toSubjectId = (
id: string,
explicit: string | undefined,
existing: string | undefined,
) =>
Effect.gen(function* () {
if (explicit !== undefined) return explicit;
if (existing !== undefined) return existing;
return rfc1035(
yield* createPhysicalName({
id,
maxLength: MAX_NAME_LENGTH,
lowercase: true,
}),
"subject",
);
});
export const parseName = (name: string, collection: string) => {
const parts = name.split("/").filter((part) => part.length > 0);
const collectionAt = parts.lastIndexOf(collection);
const locationsAt = parts.lastIndexOf("locations");
const projectsAt = parts.lastIndexOf("projects");
return {
project: projectsAt >= 0 && parts[projectsAt + 1] ? parts[projectsAt + 1]! : "",
// API resource names always carry a `locations/{location}` segment.
location: locationsAt >= 0 && parts[locationsAt + 1] ? parts[locationsAt + 1]! : "",
id:
collectionAt >= 0 && parts[collectionAt + 1]
? parts.slice(collectionAt + 1).join("/")
: lastSegment(name),
parent:
collectionAt > 0
? parts.slice(0, collectionAt).join("/")
: parts.slice(0, Math.max(0, parts.length - 1)).join("/"),
};
};
export const expandParent = (
value: string,
project: string,
location: string,
collection: string,
) => {
if (value.includes("/")) return value.replace(/\/+$/, "");
return `projects/${project}/locations/${location}/${collection}/${value}`;
};
export const userLabels = (
labels: Record<string, string | undefined> | null | undefined,
): Record<string, string> => stripInternalLabels(tagRecord(labels));
export const hasAlchemyLabelMap = (labels: Record<string, string | undefined> | null | undefined) =>
Object.keys(labels ?? {}).some((key) => key.startsWith("alchemy-"));
export const stringMapOf = (
value: Record<string, string | undefined> | null | undefined,
): Record<string, string> =>
Object.fromEntries(
Object.entries(value ?? {}).filter(
(entry): entry is [string, string] => entry[1] !== undefined,
),
);
export const sameText = (left: string | undefined, right: string | undefined) =>
(left ?? "") === (right ?? "");
export const fingerprint = (value: unknown): string => JSON.stringify(value ?? null);
export const fieldMask = (fields: Array<string | false | undefined>) =>
fields.filter((field): field is string => typeof field === "string").join(",");
export const asCount = (value: number | string | undefined, fallback: number) => {
if (value === undefined) return fallback;
const parsed = typeof value === "number" ? value : Number(value);
return Number.isFinite(parsed) ? parsed : fallback;
};
export const asCountString = (value: number | string | undefined, fallback: number) =>
String(asCount(value, fallback));
export const replaceOnIdentity = (input: {
previousId?: string;
nextId?: string;
previousLocation?: string;
nextLocation?: string;
previousParent?: string;
nextParent?: string;
extra?: boolean;
}) => {
const parentChanged =
(input.previousParent ?? "") !== "" &&
(input.nextParent ?? "") !== "" &&
(input.previousParent ?? "") !== (input.nextParent ?? "");
const locationChanged =
(input.previousLocation ?? "") !== "" &&
(input.nextLocation ?? "") !== "" &&
(input.previousLocation ?? "") !== (input.nextLocation ?? "");
const idChanged =
input.previousId !== undefined &&
input.nextId !== undefined &&
input.previousId !== input.nextId;
if (!(input.extra === true || parentChanged || locationChanged || idChanged)) {
return undefined;
}
const samePhysical =
!parentChanged &&
!locationChanged &&
input.previousId !== undefined &&
input.nextId === input.previousId;
return {
action: "replace" as const,
deleteFirst: samePhysical,
};
};
const READY_STATES = new Set(["ACTIVE", "RUNNING"]);
const FAILED_STATES = new Set(["FAILED", "ERROR"]);
export const isReadyState = (state: string | undefined) =>
READY_STATES.has((state ?? "").toUpperCase());
export const isFailedState = (state: string | undefined) =>
FAILED_STATES.has((state ?? "").toUpperCase());
/**
* Wait for a Managed Kafka operation. Cluster creates take 20–30 minutes;
* Connect clusters are similar. An operation that finished with
* `ALREADY_EXISTS` (6) is a lost create race; `notFoundOk` also accepts
* `NOT_FOUND` (5).
*/
export const waitForOperation = (operation: kafka.Operation, options?: { notFoundOk?: boolean }) =>
waitForGcpOperation(
operation,
(name) =>
kafka
.getProjectsLocationsOperations({ name })
.pipe(
Effect.catchTag("NotFound", (error) =>
options?.notFoundOk === true
? Effect.succeed<kafka.Operation>({ name, done: true })
: Effect.fail(error),
),
),
{ budget: "45 minutes" },
).pipe(
Effect.catchIf(
(error) =>
error._tag === "GCP.OperationFailed" &&
(error.code === 6 || (options?.notFoundOk === true && error.code === 5)),
() => Effect.void,
),
Effect.asVoid,
);
export const waitUntilExists = <A, E extends { readonly _tag: string }, R>(
get: Effect.Effect<A | undefined, E, R>,
name: string,
) =>
get.pipe(
Effect.filterOrFail(
(value): value is A => value !== undefined,
() => new ResourceNotResolved({ resourceName: name }),
),
Effect.retry({
while: (error) => error._tag === "GCP.ManagedKafka.ResourceNotResolved",
times: 8,
schedule: Schedule.spaced("1 second"),
}),
);
export const waitUntilGone = <A, E extends { readonly _tag: string }, R>(
get: Effect.Effect<A | undefined, E, R>,
name: string,
) =>
get.pipe(
Effect.flatMap((value) =>
value === undefined
? Effect.void
: Effect.fail(new ResourceStillExists({ resourceName: name })),
),
Effect.retry({
while: (error) => error._tag === "GCP.ManagedKafka.ResourceStillExists",
times: 10,
schedule: Schedule.spaced("2 seconds"),
}),
);
export const waitUntilReady = <A, E extends { readonly _tag: string }, R>(
get: Effect.Effect<A | undefined, E, R>,
name: string,
stateOf: (value: A) => string | undefined,
) =>
get.pipe(
Effect.filterOrFail(
(value): value is A => value !== undefined,
() => new ResourceNotResolved({ resourceName: name }),
),
Effect.filterOrFail(
(value) => !isFailedState(stateOf(value) ?? ""),
(value) => new ResourceFailed({ resourceName: name, state: stateOf(value) ?? "" }),
),
Effect.filterOrFail(
(value) => {
const state = stateOf(value) ?? "";
return isReadyState(state) || state.length === 0;
},
(value) => new ResourceNotReady({ resourceName: name, state: stateOf(value) ?? "" }),
),
Effect.retry({
while: (error) =>
error._tag === "GCP.ManagedKafka.ResourceNotReady" ||
error._tag === "GCP.ManagedKafka.ResourceNotResolved",
times: 10,
schedule: Schedule.spaced("8 seconds"),
}),
);
export const collectPages = <Page, A, E, R>(
pages: Stream.Stream<Page, E, R>,
items: (page: Page) => readonly A[] | undefined,
) =>
pages.pipe(
Stream.flatMap((page) => Stream.fromIterable(items(page) ?? [])),
Stream.runCollect,
Effect.map((chunk) => Array.from(chunk)),
);
export const listAtLocation = <A, E, R>(
project: string,
_region: string,
list: (parent: string) => Effect.Effect<A[], E, R>,
) => list(`projects/${project}/locations/-`);
export const listClusters = (project: string, region: string) =>
listAtLocation(project, region, (parent) =>
collectPages(
kafka.listProjectsLocationsClusters.pages({ parent, pageSize: 1000 }),
(page) => page.clusters,
).pipe(Effect.catchTag("NotFound", () => Effect.succeed([] as kafka.Cluster[]))),
);
export const listAlchemyClusters = (project: string, region: string) =>
listClusters(project, region).pipe(
Effect.map((clusters: kafka.Cluster[]) =>
clusters.filter((cluster) => hasAlchemyLabelMap(cluster.labels)),
),
);
export const listConnectClusters = (project: string, region: string) =>
listAtLocation(project, region, (parent) =>
collectPages(
kafka.listProjectsLocationsConnectClusters.pages({
parent,
pageSize: 1000,
}),
(page) => page.connectClusters,
).pipe(Effect.catchTag("NotFound", () => Effect.succeed([] as kafka.ConnectCluster[]))),
);
export const listAlchemyConnectClusters = (project: string, region: string) =>
listConnectClusters(project, region).pipe(
Effect.map((clusters: kafka.ConnectCluster[]) =>
clusters.filter((cluster) => hasAlchemyLabelMap(cluster.labels)),
),
);
export const listSchemaRegistries = (project: string, location: string) => {
// The `locations/-` wildcard hangs for Schema Registry; list the region.
const parents = [`projects/${project}/locations/${location}`];
return Effect.forEach(
parents,
(parent) =>
kafka
.listProjectsLocationsSchemaRegistries({
parent,
view: "SCHEMA_REGISTRY_VIEW_FULL",
})
.pipe(
Effect.map((page) => page.schemaRegistries ?? []),
Effect.catchTag(["NotFound", "SchemaRegistryRequiresCluster"], () =>
Effect.succeed([] as kafka.SchemaRegistry[]),
),
),
{ concurrency: 1 },
).pipe(
Effect.map((groups) => {
const seen = new Set<string>();
const out: kafka.SchemaRegistry[] = [];
for (const registry of groups.flat()) {
const name = registry.name ?? "";
if (name.length === 0 || seen.has(name)) continue;
seen.add(name);
out.push(registry);
}
return out;
}),
);
};
export const encodeOwnership = (labels: Record<string, string>) =>
`[alchemy ${alchemyLabelKeys.stack}=${labels[alchemyLabelKeys.stack]} ${alchemyLabelKeys.stage}=${labels[alchemyLabelKeys.stage]} ${alchemyLabelKeys.id}=${labels[alchemyLabelKeys.id]}]`;
export const parseOwnership = (text: string | undefined) => {
if (!text?.includes("[alchemy ")) return {} as Record<string, string>;
const start = text.indexOf("[alchemy ");
const end = text.indexOf("]", start);
if (end < 0) return {} as Record<string, string>;
const labels: Record<string, string> = {};
for (const part of text.slice(start + "[alchemy ".length, end).split(/\s+/)) {
const eq = part.indexOf("=");
if (eq > 0) labels[part.slice(0, eq)] = part.slice(eq + 1);
}
return labels;
};
export const hasOwnershipMarker = (text: string | undefined) =>
Object.keys(parseOwnership(text)).some((key) => key.startsWith("alchemy-"));
export const ownershipSchema = (labels: Record<string, string>) =>
JSON.stringify({
type: "record",
name: "AlchemyOwnership",
namespace: "alchemy",
doc: encodeOwnership(labels),
fields: [{ name: "ok", type: "string" }],
});
export const parseHttpJson = (body: kafka.HttpBody | undefined) =>
Effect.sync(() => {
const data = body?.data;
if (data === undefined || data.length === 0) return undefined;
try {
return JSON.parse(data) as unknown;
} catch {
/* base64 */
}
try {
return JSON.parse(Buffer.from(data, "base64").toString("utf8")) as unknown;
} catch {
return undefined;
}
});
export const subjectParent = (schemaRegistry: string, subject: string, context?: string) =>
context !== undefined && context.length > 0
? `${schemaRegistry}/contexts/${context}/subjects/${subject}`
: `${schemaRegistry}/subjects/${subject}`;
export const versionName = (parent: string, version: number | string) =>
`${parent}/versions/${version}`;
export const retryTransient = <A, E extends { readonly _tag: string }, R>(
effect: Effect.Effect<A, E, R>,
) =>
effect.pipe(
Effect.retry({
while: (error) => isTransientGcpError(error) || error._tag === "NotFound",
times: 8,
schedule: Schedule.exponential("250 millis"),
}),
);
/**
* Cluster and Connect cluster creates are limited to one request per minute
* per region, so a burst of creates is throttled with `TooManyRequests`
* until the minute window rolls over. Retry for up to ~4 minutes.
*/
export const retryCreateQuota = <A, E extends { readonly _tag: string }, R>(
effect: Effect.Effect<A, E, R>,
) =>
effect.pipe(
Effect.retry({
while: (error) => error._tag === "TooManyRequests",
times: 12,
schedule: Schedule.spaced("20 seconds"),
}),
);
export const getCluster = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsClusters({ name })
.pipe(Effect.catchTag("NotFound", () => Effect.undefined));
export const getConnectCluster = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsConnectClusters({ name })
.pipe(Effect.catchTag("NotFound", () => Effect.undefined));
export const getAcl = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsClustersAcls({ name })
.pipe(Effect.catchTag(["NotFound", "AclClusterNotFound"], () => Effect.undefined));
export const getTopic = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsClustersTopics({ name })
.pipe(Effect.catchTag("NotFound", () => Effect.undefined));
export const getConnector = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsConnectClustersConnectors({ name })
.pipe(Effect.catchTag("NotFound", () => Effect.undefined));
export const getSchemaRegistry = (name: string) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsSchemaRegistries({ name })
.pipe(
Effect.catchTag(["NotFound", "SchemaRegistryRequiresCluster"], () => Effect.undefined),
);
export const getSchemaVersion = (name: string, deleted = false) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsSchemaRegistriesSubjectsVersions({
name,
deleted,
})
.pipe(
Effect.catchTag(
["NotFound", "SchemaRegistryRequiresCluster", "SchemaRegistryPathNotFound"],
() => Effect.undefined,
),
);
export const getContextSchemaVersion = (name: string, deleted = false) =>
name.length === 0
? Effect.undefined
: kafka
.getProjectsLocationsSchemaRegistriesContextsSubjectsVersions({
name,
deleted,
})
.pipe(
Effect.catchTag(
["NotFound", "SchemaRegistryRequiresCluster", "SchemaRegistryPathNotFound"],
() => Effect.undefined,
),
);
export const stampSchemaRegistryOwnership = (
schemaRegistry: string,
labels: Record<string, string>,
) =>
Effect.gen(function* () {
const parent = subjectParent(schemaRegistry, OWNERSHIP_SUBJECT);
const schema = ownershipSchema(labels);
const latest = yield* getSchemaVersion(versionName(parent, "latest"));
if (hasOwnershipMarker(latest?.schema) && latest?.schema === schema) {
return;
}
yield* kafka
.createProjectsLocationsSchemaRegistriesSubjectsVersions({
parent,
body: {
schemaType: "AVRO",
schema,
},
})
.pipe(Effect.catchTag(["Conflict", "BadRequest"], () => Effect.void));
});
export const schemaRegistryOwnership = (schemaRegistry: string) =>
getSchemaVersion(versionName(subjectParent(schemaRegistry, OWNERSHIP_SUBJECT), "latest")).pipe(
Effect.map((version) => parseOwnership(version?.schema)),
);
export const hasSchemaRegistryOwnership = Effect.fn(function* (id: string, schemaRegistry: string) {
const expected = yield* createInternalLabels(id);
const observed = yield* schemaRegistryOwnership(schemaRegistry);
return (
observed[alchemyLabelKeys.stack] === expected[alchemyLabelKeys.stack] &&
observed[alchemyLabelKeys.stage] === expected[alchemyLabelKeys.stage] &&
observed[alchemyLabelKeys.id] === expected[alchemyLabelKeys.id]
);
});