Files
OmniRoute/tests/unit/combo-rr-fallback-advance-948.test.ts
Diego Rodrigues de Sa e Souza 3d4f3e4960 test(infra): retry recursive temp-dir removal instead of failing a shard on ENOTEMPTY (#11966) (#11968)
* test(infra): retry recursive temp-dir removal instead of failing a shard on ENOTEMPTY (#11966)

Two shards on release/v3.8.51 went red in one day with the same signature —
"ENOTEMPTY, Directory not empty: /tmp/omniroute-<test>-XXXXXX" — from
combo-same-provider-cascade (Unit Tests fast-path 4/4, on a PR that touches only
.github/) and auth-policy-embeddings-webfetch-7785 (the 20k-test TIA step). Both pass
alone and on re-run: the cleanup races something still writing into the directory
(SQLite WAL/-shm checkpoint, a worker, the backup) and under a loaded hosted runner
the window opens. 1154 test files do their own cleanup with
fs.rmSync(dir, { recursive: true, force: true }); 57 already asked for retries.

One-shot codemod (scripts/ad-hoc/codemod-rm-maxretries.mjs, kept for the record):
every rm / rmSync / rmdirSync option object with `recursive: true` and no
`maxRetries` gains `maxRetries: 5, retryDelay: 100` — Node itself then retries
ENOTEMPTY/EBUSY/EPERM for up to ~0.5 s before giving up. 2243 call sites in 1292
files under tests/, the shared tests/_setup/isolateDataDir.ts exit hook included.
Only the option object changes: no call site, assertion or import is touched.

Validation: prettier and ESLint (with the frozen suppressions) clean on all 1292
files; a random 20-file sample runs green (quota-redis-store hangs identically on
the untouched tree — it needs a Redis on localhost, an environment matter). The
four unit shards on this PR are the full run.

* fix(quality): let check-forgotten-sibling-tests read a 1,000-file diff

The gate shells out to `git diff` through execFileSync with Node's default 1 MB
maxBuffer; the 1,292-file codemod in this PR is the first diff large enough to
overflow it, and the gate died with `spawnSync git ENOBUFS` before comparing
anything. 64 MB is far above any real PR and costs nothing when unused.
2026-08-29 01:17:40 -03:00

129 lines
4.2 KiB
TypeScript

/**
* Regression for upstream 9router#948 — round-robin combo pointer must advance
* past the model that ACTUALLY served, not the eagerly-scheduled start index.
*
* With `stickyLimit: 1` ("true round-robin, one request per model"), when the
* scheduled model fails and a *different* model serves via fallback, the counter
* was advanced by +1 from the scheduled start index (eagerly, before the loop),
* so the next request started at — and reused — the fallback-served model. This
* silently degraded round-robin into hot-spotting on whichever model was healthy.
*
* This drives the REAL handleComboChat with session-stickiness disabled (to
* isolate the pure rotation pointer) and asserts two consecutive requests do NOT
* serve the same connection when the scheduled target keeps failing.
*/
import test from "node:test";
import assert from "node:assert/strict";
import fs from "node:fs";
import os from "node:os";
import path from "node:path";
const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rr-948-"));
const ORIGINAL_DATA_DIR = process.env.DATA_DIR;
process.env.DATA_DIR = TEST_DATA_DIR;
const { handleComboChat } = await import("../../open-sse/services/combo.ts");
const rrState = await import("../../open-sse/services/combo/rrState.ts");
const dbCore = await import("../../src/lib/db/core.ts");
function makeLog() {
return { info() {}, warn() {}, debug() {}, error() {} };
}
// Flat round-robin combo. conn-A always fails (fallback-eligible 429); B and C succeed.
function rrCombo(name: string) {
return {
name,
strategy: "round-robin",
// disableSessionStickiness isolates the round-robin pointer from the #3825
// per-conversation pin; stickyLimit defaults to 1 (true round-robin).
config: { maxRetries: 0, disableSessionStickiness: true },
models: [
{
kind: "model",
provider: "codex",
providerId: "codex",
model: "m-a",
connectionId: "conn-A",
id: `${name}-0`,
},
{
kind: "model",
provider: "codex",
providerId: "codex",
model: "m-b",
connectionId: "conn-B",
id: `${name}-1`,
},
{
kind: "model",
provider: "glm-cn",
providerId: "glm-cn",
model: "m-c",
connectionId: "conn-C",
id: `${name}-2`,
},
],
};
}
async function dispatchServedConnection(combo: Record<string, unknown>): Promise<string> {
let served = "?";
await handleComboChat({
body: { model: combo.name, messages: [{ role: "user", content: "hi" }], stream: false },
combo,
allCombos: [combo],
isModelAvailable: async () => true,
relayOptions: undefined,
signal: undefined,
settings: {},
log: makeLog(),
handleSingleModel: async (
_b: unknown,
modelStr: string,
target?: { connectionId?: string | null }
) => {
const conn = target?.connectionId ?? "?";
// conn-A always fails with a fallback-eligible status so rotation must fall through.
if (conn === "conn-A") {
return new Response(JSON.stringify({ error: { message: "rate limited" } }), {
status: 429,
});
}
served = conn;
return Response.json({ choices: [{ message: { role: "assistant", content: modelStr } }] });
},
});
return served;
}
test.beforeEach(() => {
rrState.rrCounters.clear();
rrState.rrStickyTargets.clear();
});
test.after(() => {
try {
dbCore.resetDbInstance?.();
} catch {
/* ignore */
}
if (ORIGINAL_DATA_DIR === undefined) delete process.env.DATA_DIR;
else process.env.DATA_DIR = ORIGINAL_DATA_DIR;
fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 });
});
test("#948: two consecutive requests do not reuse the fallback-served model", async () => {
const combo = rrCombo("rr948");
const first = await dispatchServedConnection(combo);
const second = await dispatchServedConnection(combo);
assert.notEqual(first, "?", "first request must be served by a healthy model");
assert.notEqual(second, "?", "second request must be served by a healthy model");
assert.notEqual(
first,
second,
`round-robin must advance past the served model — got ${first} twice (hot-spotting)`
);
});