solstice-ci/crates/runner-integration/src/streamer.rs

188 lines
5 KiB
Rust
Raw Normal View History

use std::sync::Arc;
use std::time::Duration;
use tracing::{debug, warn};
use uuid::Uuid;
use crate::connect::ConnectClient;
use crate::proto::runner::v1::{LogRow, UpdateLogRequest};
use crate::state::RunnerState;
const POLL_INTERVAL: Duration = Duration::from_secs(3);
/// Streams logs from logs-service to Forgejo while a job is in-flight.
/// Runs until the stop signal is received (job completed).
/// Returns the final log index so the reporter can continue from there.
pub async fn stream_logs(
client: Arc<ConnectClient>,
state: Arc<RunnerState>,
request_id: Uuid,
task_id: i64,
logs_base: String,
mut stop: tokio::sync::watch::Receiver<bool>,
) -> i64 {
let http = reqwest::Client::new();
let mut log_index: i64 = 0;
let mut last_total: i64 = 0;
loop {
// Check if we should stop
if *stop.borrow() {
// Do one final poll to catch any remaining logs
log_index = poll_and_send(
&client,
&state,
&http,
&logs_base,
request_id,
task_id,
log_index,
&mut last_total,
)
.await;
break;
}
log_index = poll_and_send(
&client,
&state,
&http,
&logs_base,
request_id,
task_id,
log_index,
&mut last_total,
)
.await;
// Wait for the next poll interval or stop signal
tokio::select! {
_ = tokio::time::sleep(POLL_INTERVAL) => {}
_ = stop.changed() => {
// Signal changed — do one more poll then exit
log_index = poll_and_send(
&client, &state, &http, &logs_base,
request_id, task_id, log_index, &mut last_total,
).await;
break;
}
}
}
debug!(task_id, log_index, "log streamer stopped");
log_index
}
/// Log category summary from logs-service.
#[derive(serde::Deserialize)]
struct LogCategorySummary {
category: String,
count: i64,
}
async fn poll_and_send(
client: &ConnectClient,
state: &RunnerState,
http: &reqwest::Client,
logs_base: &str,
request_id: Uuid,
task_id: i64,
current_index: i64,
last_total: &mut i64,
) -> i64 {
// Get total log count across all categories
let categories_url = format!(
"{}/jobs/{}/logs",
logs_base.trim_end_matches('/'),
request_id
);
let categories = match http.get(&categories_url).send().await {
Ok(resp) if resp.status().is_success() => resp
.json::<Vec<LogCategorySummary>>()
.await
.unwrap_or_default(),
_ => return current_index,
};
let new_total: i64 = categories.iter().map(|c| c.count).sum();
if new_total <= *last_total {
return current_index; // No new logs
}
*last_total = new_total;
// Fetch all logs and send new ones
// We re-fetch everything but only send lines from current_index onward.
// This is simple but not optimal for large logs — good enough for streaming.
let mut all_lines: Vec<String> = Vec::new();
for cat in &categories {
let url = format!(
"{}/jobs/{}/logs/{}",
logs_base.trim_end_matches('/'),
request_id,
cat.category
);
if let Ok(resp) = http.get(&url).send().await {
if resp.status().is_success() {
if let Ok(text) = resp.text().await {
all_lines.push(format!("::group::{}", cat.category));
for line in text.lines() {
all_lines.push(line.to_string());
}
all_lines.push("::endgroup::".to_string());
}
}
}
}
let total_lines = all_lines.len() as i64;
if total_lines <= current_index {
return current_index;
}
// Send only new lines
let new_lines = &all_lines[current_index as usize..];
if new_lines.is_empty() {
return current_index;
}
let now = prost_types::Timestamp {
seconds: time::OffsetDateTime::now_utc().unix_timestamp(),
nanos: 0,
};
let rows: Vec<LogRow> = new_lines
.iter()
.map(|line| LogRow {
time: Some(now.clone()),
content: line.clone(),
})
.collect();
let req = UpdateLogRequest {
task_id,
index: current_index,
rows,
no_more: false,
};
match client
.update_log(&req, &state.identity.uuid, &state.identity.token)
.await
{
Ok(resp) => {
debug!(
task_id,
new_lines = new_lines.len(),
ack_index = resp.ack_index,
"streamed logs"
);
resp.ack_index
}
Err(e) => {
warn!(error = %e, task_id, "failed to stream logs");
current_index
}
}
}