lemmy/src/prometheus_metrics.rs

123 lines
3.3 KiB
Rust
Raw Normal View History

// TODO: should really not unwrap everywhere here....
#![allow(clippy::unwrap_used)]
use actix_web::{rt::System, web, App, HttpResponse, HttpServer, Responder};
use lemmy_api_common::context::LemmyContext;
use lemmy_utils::settings::structs::PrometheusConfig;
use prometheus::{default_registry, Encoder, Gauge, Opts, TextEncoder};
use std::{
net::{IpAddr, Ipv4Addr},
sync::Arc,
thread,
};
struct PromContext {
lemmy: LemmyContext,
db_pool_metrics: DbPoolMetrics,
}
struct DbPoolMetrics {
max_size: Gauge,
size: Gauge,
available: Gauge,
}
static DEFAULT_BIND: IpAddr = IpAddr::V4(Ipv4Addr::new(127, 0, 0, 1));
static DEFAULT_PORT: i32 = 10002;
pub fn serve_prometheus(config: Option<&PrometheusConfig>, lemmy_context: LemmyContext) {
let context = Arc::new(PromContext {
lemmy: lemmy_context,
db_pool_metrics: create_db_pool_metrics(),
});
let (bind, port) = match config {
Some(config) => (
config.bind.unwrap_or(DEFAULT_BIND),
config.port.unwrap_or(DEFAULT_PORT),
),
None => (DEFAULT_BIND, DEFAULT_PORT),
};
// spawn thread that blocks on handling requests
// only mapping /metrics to a handler
thread::spawn(move || {
let sys = System::new();
sys.block_on(async {
let server = HttpServer::new(move || {
App::new()
.app_data(web::Data::new(Arc::clone(&context)))
.route("/metrics", web::get().to(metrics))
})
.bind((bind, port as u16))
2023-07-10 12:04:39 +00:00
.unwrap_or_else(|_| panic!("Cannot bind to {}:{}", bind, port))
.run();
if let Err(err) = server.await {
eprintln!("Prometheus server error: {}", err);
}
})
});
}
// handler for the /metrics path
async fn metrics(context: web::Data<Arc<PromContext>>) -> impl Responder {
// collect metrics
collect_db_pool_metrics(&context).await;
let mut buffer = Vec::new();
let encoder = TextEncoder::new();
// gather metrics from registry and encode in prometheus format
let metric_families = prometheus::gather();
encoder.encode(&metric_families, &mut buffer).unwrap();
let output = String::from_utf8(buffer).unwrap();
HttpResponse::Ok().body(output)
}
// create lemmy_db_pool_* metrics and register them with the default registry
fn create_db_pool_metrics() -> DbPoolMetrics {
let metrics = DbPoolMetrics {
max_size: Gauge::with_opts(Opts::new(
"lemmy_db_pool_max_connections",
"Maximum number of connections in the pool",
))
.unwrap(),
size: Gauge::with_opts(Opts::new(
"lemmy_db_pool_connections",
"Current number of connections in the pool",
))
.unwrap(),
available: Gauge::with_opts(Opts::new(
"lemmy_db_pool_available_connections",
"Number of available connections in the pool",
))
.unwrap(),
};
default_registry()
.register(Box::new(metrics.max_size.clone()))
.unwrap();
default_registry()
.register(Box::new(metrics.size.clone()))
.unwrap();
default_registry()
.register(Box::new(metrics.available.clone()))
.unwrap();
metrics
}
async fn collect_db_pool_metrics(context: &PromContext) {
Make functions work with both connection and pool (#3420) * a lot * merge * Fix stuff broken by merge * Get rid of repetitive `&mut *context.conn().await?` * Add blank lines under each line with `conn =` * Fix style mistakes (partial) * Revert "Fix style mistakes (partial)" This reverts commit 48a033b87f4fdc1ce14ff86cc019e1c703cd2741. * Revert "Add blank lines under each line with `conn =`" This reverts commit 773a6d3beba2cf89eac75913078b40c4f5190dd4. * Revert "Get rid of repetitive `&mut *context.conn().await?`" This reverts commit d2c6263ea13710177d49b2791278db5ad115fca5. * Use DbConn for CaptchaAnswer methods * DbConn trait * Remove more `&mut *` * Fix stuff * Re-run CI * try to make ci start * fix * fix * Fix api_common::utils * Fix apub::activities::block * Fix apub::api::resolve_object * Fix some things * Revert "Fix some things" This reverts commit 2bf8574bc8333d8d34ca542d61a0a5b50039c24d. * Revert "Fix apub::api::resolve_object" This reverts commit 3e4059aabbe485b2ff060bdeced8ef958ff62832. * Revert "Fix apub::activities::block" This reverts commit 3b02389abd780a7b1b8a2c89e26febdaa6a12159. * Revert "Fix api_common::utils" This reverts commit 7dc73de613a5618fa57eb06450f3699bbcb41254. * Revert "Revert "Fix api_common::utils"" This reverts commit f740f115e5457e83e53cc223e48196a2c47a9975. * Revert "Revert "Fix apub::activities::block"" This reverts commit 2ee206af7c885c10092cf209bf4a5b1d60327866. * Revert "Revert "Fix apub::api::resolve_object"" This reverts commit 96ed8bf2e9dcadae760743929498312334e23d2e. * Fix fetch_local_site_data * Fix get_comment_parent_creator * Remove unused perma deleted text * Fix routes::feeds * Fix lib.rs * Update lib.rs * rerun ci * Attempt to create custom GetConn and RunQueryDsl traits * Start over * Add GetConn trait * aaaa * Revert "aaaa" This reverts commit acc9ca1aed10c39efdd91cefece066e035a1fe80. * Revert "Revert "aaaa"" This reverts commit 443a2a00a56d152bb7eb429efd0d29a78e21b163. * still aaaaaaaaaaaaa * Return to earlier thing Revert "Add GetConn trait" This reverts commit ab4e94aea5bd9d34cbcddf017339131047e75344. * Try to use DbPool enum * Revert "Try to use DbPool enum" This reverts commit e4d1712646a52006b865a1fbe0dcf79976fdb027. * DbConn and DbPool enums (db_schema only fails to compile for tests) * fmt * Make functions take `&mut DbPool<'_>` and make db_schema tests compile * Add try_join_with_pool macro and run fix-clippy on more crates * Fix some errors * I did it * Remove function variants that take connection * rerun ci * rerun ci * rerun ci
2023-07-11 13:09:59 +00:00
let pool_status = context.lemmy.inner_pool().status();
context
.db_pool_metrics
.max_size
.set(pool_status.max_size as f64);
context.db_pool_metrics.size.set(pool_status.size as f64);
context
.db_pool_metrics
.available
.set(pool_status.available as f64);
}