wire the authenticated OTLP push into vcs_metrics and hive-jobq-metrics

This commit is contained in:
damocles 2026-08-26 22:56:44 +02:00 committed by mara
commit 361396cab8
6 changed files with 147 additions and 10 deletions

3
Cargo.lock generated
View file

@ -1794,9 +1794,11 @@ name = "hive-jobq-metrics"
version = "0.1.0"
dependencies = [
"anyhow",
"async-trait",
"hive-jobq",
"hive-jobq-wire",
"opentelemetry",
"opentelemetry-http",
"opentelemetry-otlp",
"opentelemetry_sdk",
"tokio",
@ -4583,6 +4585,7 @@ dependencies = [
"hive-types",
"hmac 0.13.0",
"opentelemetry",
"opentelemetry-http",
"opentelemetry-otlp",
"opentelemetry_sdk",
"problem_details",

View file

@ -9,10 +9,18 @@ workspace = true
[dependencies]
anyhow.workspace = true
# `spawn_exporter`'s optional `http_client` param: an authenticated push
# (`swarm-controller`'s case, wired through `swarm_queue_client::
# otel_http_client`) needs a caller-supplied `HttpClient` per request, not
# a static header. This crate stays generic over that caller's concrete
# client type — see `BoxedHttpClient`'s own doc — so it depends on the
# trait, never on `swarm-queue-client` itself.
async-trait.workspace = true
hive-jobq.workspace = true
hive-jobq-wire.workspace = true
opentelemetry.workspace = true
opentelemetry_sdk.workspace = true
opentelemetry-otlp.workspace = true
opentelemetry-http.workspace = true
tokio.workspace = true
tracing.workspace = true

View file

@ -23,13 +23,38 @@ use std::sync::{Arc, Mutex};
use std::time::Duration;
use anyhow::{Context, Result};
use async_trait::async_trait;
use hive_jobq_wire::StateCount;
use opentelemetry::KeyValue;
use opentelemetry::metrics::MeterProvider as _;
use opentelemetry_otlp::{MetricExporter, Protocol, WithExportConfig};
use opentelemetry_http::{Bytes, HttpClient, HttpError, Request, Response};
use opentelemetry_otlp::{MetricExporter, Protocol, WithExportConfig, WithHttpConfig};
use opentelemetry_sdk::Resource;
use opentelemetry_sdk::metrics::{PeriodicReader, SdkMeterProvider};
/// Type-erases a caller's concrete [`HttpClient`] impl so [`spawn_exporter`]
/// can accept "any client, or none" without this crate depending on any one
/// caller's concrete type (`swarm-controller`'s
/// `swarm_queue_client::otel_http_client::AuthenticatedHttpClient` today,
/// conceivably something else from a future caller) — the same genericity
/// this crate already has over `N`/`R`, applied to the one other caller-
/// supplied thing it touches.
///
/// A thin delegating newtype rather than a blanket `impl HttpClient for
/// Box<dyn HttpClient>`: the orphan rule refuses that blanket impl outright
/// (`Box` and `dyn HttpClient` are both foreign to this crate), so a local
/// wrapper is the standard way around it, not a workaround chosen over a
/// cleaner alternative.
#[derive(Debug)]
struct BoxedHttpClient(Box<dyn HttpClient>);
#[async_trait]
impl HttpClient for BoxedHttpClient {
async fn send_bytes(&self, request: Request<Bytes>) -> Result<Response<Bytes>, HttpError> {
self.0.send_bytes(request).await
}
}
/// Default export cadence when `HYPERHIVE_OTEL_METRIC_INTERVAL_MS` is unset.
/// Matches `hive-c0re::stats::otel_metrics`'s default — no reason for the
/// exporters across this tree to disagree on how fresh "current" means by
@ -53,9 +78,18 @@ const DEFAULT_INTERVAL: Duration = Duration::from_mins(1);
/// Generic over `N`/`R` (the same parameters [`hive_jobq::scheduler::Scheduler`]
/// itself takes) rather than any one host's node/resource vocabulary — the
/// whole point of this crate is a shape any `hive-jobq` host can plug in.
///
/// `http_client`: `None` builds the exporter with its default (endpoint-
/// only, no per-request auth) transport — the shape every caller used
/// before this parameter existed. `Some(client)` routes every export
/// through that [`HttpClient`] instead, for a caller whose destination
/// checks a bearer token per request rather than a static header
/// (`swarm-controller`'s case — see `swarm_queue_client::otel_http_client`'s
/// module doc for why a static header doesn't fit an expiring token).
pub fn spawn_exporter<N, R>(
jobq: Arc<Mutex<hive_jobq::scheduler::Scheduler<N, R>>>,
service_name: &str,
http_client: Option<Box<dyn HttpClient>>,
) -> Option<SdkMeterProvider>
where
N: Send + 'static,
@ -83,7 +117,7 @@ where
}
});
match build_provider(interval, snapshot, service_name) {
match build_provider(interval, snapshot, service_name, http_client) {
Ok(provider) => {
tracing::info!(%endpoint, ?interval, %service_name, "otel jobq-metrics: exporter enabled");
Some(provider)
@ -99,15 +133,22 @@ fn build_provider(
interval: Duration,
snapshot: Arc<Mutex<Vec<StateCount>>>,
service_name: &str,
http_client: Option<Box<dyn HttpClient>>,
) -> Result<SdkMeterProvider> {
// Same http/json, endpoint-from-env-only construction as
// `hive-c0re::stats::otel_metrics` — see that module's doc comment for
// why `with_endpoint` is deliberately never called here.
let exporter = MetricExporter::builder()
let mut builder = MetricExporter::builder()
.with_http()
.with_protocol(Protocol::HttpJson)
.build()
.context("build OTLP metric exporter")?;
.with_protocol(Protocol::HttpJson);
// Only when the caller supplied one — see `spawn_exporter`'s doc for
// who does and why. Wrapped in `BoxedHttpClient` because
// `with_http_client` wants an owned `T: HttpClient`, not the trait
// object this parameter is typed as (see that newtype's own doc).
if let Some(client) = http_client {
builder = builder.with_http_client(BoxedHttpClient(client));
}
let exporter = builder.build().context("build OTLP metric exporter")?;
let reader = PeriodicReader::builder(exporter)
.with_interval(interval)
.build();

View file

@ -65,7 +65,13 @@ hive-types.workspace = true
# `auth`'s bridge client — same crate the bridge itself uses to define the
# request/response shape, so the two ends cannot drift. `forge.rs` also
# uses this directly for `StatusCode` in its error-classification helpers.
reqwest.workspace = true
#
# `blocking` on top of the workspace default: `vcs_metrics`'s OTLP exporter
# runs on a thread with no tokio reactor (see that module's doc), so the
# `reqwest::blocking::Client` it hands to `AuthenticatedHttpClient` has to
# come from the blocking half of this crate, not the async one every other
# consumer here uses.
reqwest = { workspace = true, features = ["blocking"] }
serde.workspace = true
serde_json.workspace = true
swarm-authelia-bridge-sock.workspace = true
@ -77,7 +83,17 @@ swarm-authelia-bridge-sock.workspace = true
# `kv` for the same reason one level in: the status bucket's name and
# creation config are shared with the hive that writes it, so this end does
# not get to declare them privately.
swarm-queue-client = { workspace = true, features = ["kv"] }
#
# `otel-auth` for `vcs_metrics`'s `AuthenticatedHttpClient` — see that
# feature's own comment in `swarm-queue-client`'s Cargo.toml.
swarm-queue-client = { workspace = true, features = ["kv", "otel-auth"] }
# The `HttpClient` trait itself, so `main.rs` can name
# `Box<dyn opentelemetry_http::HttpClient>` when handing
# `vcs_metrics::authenticated_http_client()`'s output to
# `hive_jobq_metrics::spawn_exporter` — a transitive dependency (via
# `swarm-queue-client`'s `otel-auth` feature above) isn't enough to `use`
# it directly, Cargo requires a direct entry for that.
opentelemetry-http.workspace = true
tokio.workspace = true
tracing.workspace = true
tracing-subscriber.workspace = true

View file

@ -1157,8 +1157,21 @@ async fn main() -> Result<()> {
// `PeriodicReader`, so it must live as long as `main` does (which it
// does here: this binding never goes out of scope before the process
// exits via `axum::serve(...).await` below).
// Own `AuthenticatedHttpClient` instance, independent of `vcs_metrics`'s
// — both read the same env vars, and a client is cheap enough
// (one `reqwest::blocking::Client`) that sharing one across two
// otherwise-independent exporters isn't worth the plumbing. `Ok` becomes
// `Some`, `Err` becomes `None` — same "log and carry on" shape as the
// rest of this function's optional wiring; a controller that can't
// authenticate its jobq push still serves every other route.
let jobq_http_client = crate::vcs_metrics::authenticated_http_client()
.inspect_err(
|e| tracing::warn!(error = ?e, "otel jobq-metrics: no authenticated http client"),
)
.ok()
.map(|c| Box::new(c) as Box<dyn opentelemetry_http::HttpClient>);
let _jobq_metrics_provider =
hive_jobq_metrics::spawn_exporter(Arc::clone(&jobq), "swarm-controller");
hive_jobq_metrics::spawn_exporter(Arc::clone(&jobq), "swarm-controller", jobq_http_client);
// Same "log and carry on" shape as the queue/bridge/forge wiring above.
// A controller that cannot hold a webhook secret still serves every

View file

@ -25,7 +25,7 @@ use std::time::Duration;
use anyhow::{Context, Result};
use opentelemetry::KeyValue;
use opentelemetry::metrics::{Counter, MeterProvider as _};
use opentelemetry_otlp::{MetricExporter, Protocol, WithExportConfig};
use opentelemetry_otlp::{MetricExporter, Protocol, WithExportConfig, WithHttpConfig};
use opentelemetry_sdk::Resource;
use opentelemetry_sdk::metrics::{PeriodicReader, SdkMeterProvider};
@ -114,9 +114,16 @@ fn build_provider(interval: Duration) -> Result<SdkMeterProvider> {
// Same http/json, endpoint-from-env-only construction as
// `hive_jobq_metrics::build_provider` — see that module's doc comment
// for why `with_endpoint` is deliberately never called here.
//
// `.with_http_client(..)`: unlike the hive tier's own exporters, this
// one crosses a real trust boundary (the swarm-tier `otlp/swarm`
// receiver checks a bearer token's audience, see
// `swarm_queue_client::otel_http_client`'s module doc) — a plain
// endpoint-only client would be refused at the receiver.
let exporter = MetricExporter::builder()
.with_http()
.with_protocol(Protocol::HttpJson)
.with_http_client(authenticated_http_client()?)
.build()
.context("build OTLP metric exporter")?;
let reader = PeriodicReader::builder(exporter)
@ -128,6 +135,55 @@ fn build_provider(interval: Duration) -> Result<SdkMeterProvider> {
.build())
}
/// Build the [`swarm_queue_client::otel_http_client::AuthenticatedHttpClient`]
/// this exporter pushes through.
///
/// The queue identity (`SWARM_CONTROLLER_OIDC_*`) is read fresh here rather
/// than threaded in from a caller — by the time this runs, [`endpoint`] has
/// already confirmed OTEL is configured for this host, and `queueEnv` is set
/// **unconditionally** for every `swarm-controller` (the nix module's own
/// words: "a controller without a queue is not a lighter controller, it is
/// a broken one"). So an absent queue identity here is never a supported
/// "OTEL enabled, queue not configured" deployment shape — it is a
/// deployment bug, and gets the same `Err` treatment `build()`'s caller
/// already applies to any other exporter-construction failure.
///
/// `SWARM_CONTROLLER_OTEL_AUDIENCE` names the audience `otlp/swarm`'s
/// authenticator checks — set by the same nix option that sets
/// `OTEL_EXPORTER_OTLP_ENDPOINT`, so the two are never independently absent.
///
/// No CA handling here unlike [`swarm_queue_client::QueueConfig::ca_file`]
/// (which `mint_token_for_blocking` already applies to the TOKEN endpoint
/// internally): the metrics endpoint is reached through the gateway with a
/// certificate the swarm CA issued, and this whole process already trusts
/// that CA via `SSL_CERT_FILE` (`hive-ca-trust.nix`, applied host-wide to
/// this unit for the same reason the forge client needs no special
/// handling either) — a second, narrower trust bundle here would just be
/// the same fact stated twice.
pub(crate) fn authenticated_http_client()
-> Result<swarm_queue_client::otel_http_client::AuthenticatedHttpClient> {
let cfg = swarm_queue_client::QueueConfig::from_env("SWARM_CONTROLLER")
.context("reading the queue identity this daemon authenticates its OTLP push with")?
.ok_or_else(|| {
anyhow::anyhow!(
"SWARM_CONTROLLER_NATS_URL and friends are unset, but OTEL_EXPORTER_OTLP_ENDPOINT \
is the queue is required for every controller, so this combination is a \
deployment bug, not a supported partial config"
)
})?;
let audience = std::env::var("SWARM_CONTROLLER_OTEL_AUDIENCE").context(
"SWARM_CONTROLLER_OTEL_AUDIENCE is unset, but OTEL_EXPORTER_OTLP_ENDPOINT is — the nix \
module sets both together",
)?;
Ok(
swarm_queue_client::otel_http_client::AuthenticatedHttpClient::new(
reqwest::blocking::Client::new(),
cfg,
audience,
),
)
}
/// `service.name = swarm-controller` plus whatever the operator set in
/// `HYPERHIVE_OTEL_EXTRA_RESOURCE_ATTRIBUTES` — same channel
/// `hive_jobq_metrics::resource` reads, since both live in this one binary.