Add complete system qualification evidence

This commit is contained in:
ms2k 2026-07-18 14:09:04 +09:00
commit 6952767503
6 changed files with 1256 additions and 84 deletions

View file

@ -42,4 +42,9 @@ async fn deployed_api_accepts_an_authorized_job_for_coordination_rebuild() {
let body: Value = submitted.json().await.unwrap();
serde_json::from_value::<JobResourceDto>(body.clone()).unwrap();
assert_eq!(body["phase"], "Queued");
if let Ok(expected) = std::env::var("REMORA_EXPECTED_JOB_ID")
&& !expected.is_empty()
{
assert_eq!(body["id"], expected);
}
}

View file

@ -40,7 +40,9 @@ use tokio_tungstenite::{
mod support;
use support::{mint_provider_token, oidc_provider, required_env, revoke_remora_grants};
use support::{
mint_provider_token, oidc_provider, required_env, restore_remora_grants, revoke_remora_grants,
};
type Socket = WebSocketStream<MaybeTlsStream<tokio::net::TcpStream>>;
const BROWSER_ORIGIN: &str = "https://remora.test";
@ -60,6 +62,7 @@ impl Harness {
if std::env::var("REMORA_LIVE_EVENTS_REQUIRE").is_err() {
return None;
}
restore_remora_grants().await.unwrap();
let pool = connect(
&required_env("DATABASE_URL"),
PoolConfig {
@ -156,7 +159,7 @@ impl Harness {
PublicApiConfig {
cursor_secret: b"live-events-fixture-cursor-secret".to_vec(),
member_access_ttl: Duration::from_secs(60),
event_retention: Duration::from_millis(500),
event_retention: Duration::from_secs(2),
event_poll_interval: Duration::from_millis(250),
event_heartbeat_interval: Duration::from_millis(100),
event_ticket_ttl: Duration::from_secs(30),
@ -403,7 +406,7 @@ async fn live_event_replay_authorization_progress_and_resync_converge() {
.await
.unwrap();
let expiring_cursor = fresh["event_cursor"].as_str().unwrap().to_owned();
tokio::time::sleep(Duration::from_millis(600)).await;
tokio::time::sleep(Duration::from_millis(2_100)).await;
let mut expired = harness
.connect(&harness.token, Some(&expiring_cursor))
.await;
@ -749,12 +752,25 @@ async fn next_frame(socket: &mut Socket) -> EventFrameDto {
}
async fn next_job_frame(socket: &mut Socket, job_id: &str) -> EventFrameDto {
loop {
let frame = next_frame(socket).await;
if matches!(&frame, EventFrameDto::JobChanged(event) if event.job_id.as_str() == job_id) {
return frame;
tokio::time::timeout(Duration::from_secs(10), async {
loop {
let frame = next_frame(socket).await;
match &frame {
EventFrameDto::JobChanged(event) if event.job_id.as_str() == job_id => {
return frame;
}
EventFrameDto::ResyncRequired(event) => {
panic!(
"unexpected resync while waiting for job event: {:?}",
event.reason
)
}
_ => {}
}
}
}
})
.await
.expect("job event timeout")
}
async fn next_job_progress(
@ -762,42 +778,54 @@ async fn next_job_progress(
job_id: &str,
completed_bytes: Option<u64>,
) -> EventFrameDto {
loop {
let frame = next_job_frame(socket, job_id).await;
let EventFrameDto::JobChanged(event) = &frame else {
unreachable!()
};
let observed = event
.progress
.0
.as_ref()
.map(|progress| progress.completed_bytes.get());
if observed == completed_bytes {
return frame;
tokio::time::timeout(Duration::from_secs(10), async {
loop {
let frame = next_job_frame(socket, job_id).await;
let EventFrameDto::JobChanged(event) = &frame else {
unreachable!()
};
let observed = event
.progress
.0
.as_ref()
.map(|progress| progress.completed_bytes.get());
if observed == completed_bytes {
return frame;
}
}
}
})
.await
.expect("job progress timeout")
}
async fn next_deletion_frame(socket: &mut Socket, operation_id: &str) -> EventFrameDto {
loop {
let frame = next_frame(socket).await;
if matches!(
&frame,
EventFrameDto::DeletionChanged(event)
if event.operation_id.as_str() == operation_id
) {
return frame;
tokio::time::timeout(Duration::from_secs(10), async {
loop {
let frame = next_frame(socket).await;
if matches!(
&frame,
EventFrameDto::DeletionChanged(event)
if event.operation_id.as_str() == operation_id
) {
return frame;
}
}
}
})
.await
.expect("deletion event timeout")
}
async fn next_resync(socket: &mut Socket) -> EventFrameDto {
loop {
let frame = next_frame(socket).await;
if matches!(frame, EventFrameDto::ResyncRequired(_)) {
return frame;
tokio::time::timeout(Duration::from_secs(10), async {
loop {
let frame = next_frame(socket).await;
if matches!(frame, EventFrameDto::ResyncRequired(_)) {
return frame;
}
}
}
})
.await
.expect("resync event timeout")
}
fn assert_resync(frame: EventFrameDto, expected: ResyncReasonDto) {
@ -807,7 +835,7 @@ fn assert_resync(frame: EventFrameDto, expected: ResyncReasonDto) {
fn assert_resync_reason(frame: EventFrameDto, expected: ResyncReasonDto) -> u64 {
let EventFrameDto::ResyncRequired(frame) = frame else {
panic!("expected resync_required");
panic!("expected resync_required, got {frame:?}");
};
assert_eq!(frame.reason, expected);
frame.sequence.get()

View file

@ -35,7 +35,9 @@ use time::OffsetDateTime;
mod support;
use support::{mint_provider_token, oidc_provider, required_env, revoke_remora_grants};
use support::{
mint_provider_token, oidc_provider, required_env, restore_remora_grants, revoke_remora_grants,
};
struct Harness {
pool: PgPool,
@ -51,6 +53,7 @@ impl Harness {
if std::env::var("REMORA_PUBLIC_API_REQUIRE").is_err() {
return None;
}
restore_remora_grants().await.unwrap();
let pool = connect(
&required_env("DATABASE_URL"),
PoolConfig {

View file

@ -47,6 +47,14 @@ pub async fn revoke_remora_grants() -> Result<(), String> {
}
}
pub async fn restore_remora_grants() -> Result<(), String> {
match required_env("REMORA_OIDC_PROVIDER").as_str() {
"keycloak" => restore_keycloak_remora_roles().await,
"zitadel" => restore_zitadel_remora_roles().await,
provider => Err(format!("unsupported OIDC provider {provider}")),
}
}
async fn mint_keycloak_token() -> Result<String, String> {
let response = Client::new()
.post(format!(
@ -262,6 +270,103 @@ async fn revoke_keycloak_remora_roles() -> Result<(), String> {
}
}
async fn restore_keycloak_remora_roles() -> Result<(), String> {
let realm = required_env("REMORA_OIDC_TENANT");
let client_id = required_env("REMORA_OIDC_CLIENT_ID");
let username = required_env("REMORA_OIDC_USERNAME");
let client = Client::new();
let admin_response = client
.post("http://keycloak.remora.test:8080/realms/master/protocol/openid-connect/token")
.form(&[
("grant_type", "password"),
("client_id", "admin-cli"),
("username", "remora-admin"),
("password", "synthetic-keycloak"),
])
.send()
.await
.map_err(|error| error.to_string())?;
let admin_body: Value = admin_response
.json()
.await
.map_err(|error| error.to_string())?;
let admin_token = admin_body["access_token"]
.as_str()
.ok_or_else(|| "keycloak admin token missing".to_owned())?;
let users: Value = client
.get(format!(
"http://keycloak.remora.test:8080/admin/realms/{realm}/users?username={username}&exact=true"
))
.bearer_auth(admin_token)
.send()
.await
.map_err(|error| error.to_string())?
.json()
.await
.map_err(|error| error.to_string())?;
let user_id = users
.as_array()
.and_then(|rows| rows.first())
.and_then(|row| row["id"].as_str())
.ok_or_else(|| "keycloak user missing".to_owned())?;
let clients: Value = client
.get(format!(
"http://keycloak.remora.test:8080/admin/realms/{realm}/clients?clientId={client_id}"
))
.bearer_auth(admin_token)
.send()
.await
.map_err(|error| error.to_string())?
.json()
.await
.map_err(|error| error.to_string())?;
let client_uuid = clients
.as_array()
.and_then(|rows| rows.first())
.and_then(|row| row["id"].as_str())
.ok_or_else(|| "keycloak client missing".to_owned())?;
let roles: Value = client
.get(format!(
"http://keycloak.remora.test:8080/admin/realms/{realm}/clients/{client_uuid}/roles"
))
.bearer_auth(admin_token)
.send()
.await
.map_err(|error| error.to_string())?
.json()
.await
.map_err(|error| error.to_string())?;
let storage_role = required_env("REMORA_OIDC_STORAGE_ROLE");
let roles = roles
.as_array()
.cloned()
.unwrap_or_default()
.into_iter()
.filter(|role| {
role["name"]
.as_str()
.is_some_and(|name| name == "remora:v1:admin" || name == storage_role)
})
.collect::<Vec<_>>();
let response = client
.post(format!(
"http://keycloak.remora.test:8080/admin/realms/{realm}/users/{user_id}/role-mappings/clients/{client_uuid}"
))
.bearer_auth(admin_token)
.json(&roles)
.send()
.await
.map_err(|error| error.to_string())?;
if response.status().is_success() {
Ok(())
} else {
Err(format!(
"keycloak role restore failed: {}",
response.status()
))
}
}
async fn revoke_zitadel_remora_roles() -> Result<(), String> {
let issuer = required_env("REMORA_OIDC_ISSUER");
let organization = required_env("REMORA_OIDC_TENANT");
@ -316,3 +421,64 @@ async fn revoke_zitadel_remora_roles() -> Result<(), String> {
))
}
}
async fn restore_zitadel_remora_roles() -> Result<(), String> {
let issuer = required_env("REMORA_OIDC_ISSUER");
let organization = required_env("REMORA_OIDC_TENANT");
let user_id = required_env("REMORA_OIDC_USER_ID");
let project_id = required_env("REMORA_OIDC_PROJECT_ID");
let token = required_env("REMORA_OIDC_MGMT_TOKEN");
let host = issuer
.strip_prefix("http://")
.or_else(|| issuer.strip_prefix("https://"))
.unwrap_or(&issuer);
let client = Client::new();
let grants: Value = client
.post(format!("{issuer}/management/v1/users/grants/_search"))
.header("Host", host)
.bearer_auth(&token)
.header("x-zitadel-orgid", &organization)
.json(&json!({ "queries": [{ "userIdQuery": { "userId": user_id } }] }))
.send()
.await
.map_err(|error| error.to_string())?
.json()
.await
.map_err(|error| error.to_string())?;
let grant_id = grants["result"]
.as_array()
.and_then(|rows| {
rows.iter().find(|row| {
row["projectId"]
.as_str()
.is_some_and(|value| value == project_id)
})
})
.and_then(|row| row["id"].as_str())
.ok_or_else(|| "zitadel project grant missing".to_owned())?;
let response = client
.put(format!(
"{issuer}/management/v1/users/{user_id}/grants/{grant_id}"
))
.header("Host", host)
.bearer_auth(token)
.header("x-zitadel-orgid", organization)
.json(&json!({
"roleKeys": [
"test-user",
"remora:v1:admin",
required_env("REMORA_OIDC_STORAGE_ROLE")
]
}))
.send()
.await
.map_err(|error| error.to_string())?;
if response.status().is_success() {
Ok(())
} else {
Err(format!(
"zitadel role restore failed: {}",
response.status()
))
}
}

View file

@ -1,5 +1,5 @@
use std::sync::Arc;
use std::time::Duration;
use std::time::{Duration, Instant};
use remora_coordination::{
ActivateOptions, CoordinationError, CoordinationStore, LeaseOwner, LocalPermitGate,
@ -164,6 +164,85 @@ async fn reserve_contention_respects_global_limit() {
.expect("cleanup");
}
#[tokio::test]
async fn permit_contention_benchmark_preserves_global_bound() {
if std::env::var("REMORA_QUALIFICATION_REQUIRE").is_err() {
return;
}
const TASKS: usize = 16;
const OPERATIONS_PER_TASK: usize = 16;
const GLOBAL_LIMIT: u64 = 8;
let Some(store) = test_store(
"qualification-permits",
GLOBAL_LIMIT,
Duration::from_secs(30),
)
.await
else {
return;
};
let jobs = (0..TASKS * OPERATIONS_PER_TASK)
.map(|index| format!("job-qualification-{index:04}"))
.collect::<Vec<_>>();
let job_refs = jobs.iter().map(String::as_str).collect::<Vec<_>>();
store.drop_job_state(&job_refs).await.expect("clean");
let barrier = Arc::new(Barrier::new(TASKS));
let started = Instant::now();
let mut tasks = Vec::with_capacity(TASKS);
for task_index in 0..TASKS {
let store = store.clone();
let barrier = Arc::clone(&barrier);
let worker = format!("worker-qualification-{task_index:02}");
let task_jobs =
jobs[task_index * OPERATIONS_PER_TASK..(task_index + 1) * OPERATIONS_PER_TASK].to_vec();
tasks.push(tokio::spawn(async move {
let mut durations = Vec::with_capacity(OPERATIONS_PER_TASK);
barrier.wait().await;
for job in task_jobs {
let operation_started = Instant::now();
loop {
let outcome = store.reserve(&job, &worker).await.expect("reserve");
if outcome.ok() {
break;
}
assert_eq!(outcome.code, OpCode::FailFull);
tokio::task::yield_now().await;
}
assert!(store.global_used().await.expect("used") <= GLOBAL_LIMIT);
assert!(
store
.free_reserve(&job, Some(&worker))
.await
.expect("free")
.ok()
);
durations.push(operation_started.elapsed());
}
durations
}));
}
let mut durations = Vec::with_capacity(TASKS * OPERATIONS_PER_TASK);
for task in tasks {
durations.extend(task.await.expect("join"));
}
durations.sort_unstable();
let elapsed = started.elapsed();
let p50 = durations[durations.len() / 2];
let p99 = durations[durations.len() * 99 / 100];
let operations_per_second = durations.len() as f64 / elapsed.as_secs_f64();
eprintln!(
"permit benchmark operations={} elapsed_ms={} p50_us={} p99_us={} operations_per_second={operations_per_second:.1}",
durations.len(),
elapsed.as_millis(),
p50.as_micros(),
p99.as_micros()
);
assert!(elapsed < Duration::from_secs(30));
assert_eq!(store.global_used().await.expect("used"), 0);
store.drop_job_state(&job_refs).await.expect("cleanup");
}
#[tokio::test]
async fn release_and_expire_preserve_reserved_only_global_slot() {
let Some(store) = test_store("permit-reserved", 1, Duration::from_secs(10)).await else {

View file

@ -0,0 +1,18 @@
use sqlx::postgres::PgPoolOptions;
#[tokio::test]
async fn remove_standalone_credential_fixture_rows() {
if std::env::var("REMORA_QUALIFICATION_REQUIRE").is_err() {
return;
}
let database_url = std::env::var("DATABASE_URL").expect("DATABASE_URL");
let pool = PgPoolOptions::new()
.max_connections(1)
.connect(&database_url)
.await
.expect("connect postgres");
sqlx::query("TRUNCATE execution_grant_credential_pins, source_credentials CASCADE")
.execute(&pool)
.await
.expect("clear standalone credential fixtures");
}