Files
task_pilot/backend/src/runs/repository.rs
T
CNWei 815258c19b feat: 加固运行期可靠性并补齐调度与观测接口
- 运行句柄按 run_id 归属,修复取消后重跑时闸门被旧运行删除的竞态
- 调度循环改用取消令牌退出,启动时对账遗留运行,校验 rows_affected
- 三类任务统一参与调度,新增 /api/v1/stats 与 /api/v1/schedule
- SQLite 启用 WAL 与 busy_timeout,日志清理跳过活跃文件,单行上限 256 KiB
- 前端缓存管理令牌并在 403 时重试一次,仪表盘改用统计接口
2026-10-26 00:38:16 +08:00

275 lines
7.6 KiB
Rust

use chrono::{DateTime, Utc};
use sqlx::{FromRow, SqlitePool};
use crate::{
error::{ApiError, ApiResult},
runs::model::{RunListQuery, RunStatus, RunTrigger, TaskRunListResponse, TaskRunResponse},
};
const DEFAULT_PAGE_SIZE: u32 = 20;
const MAX_PAGE_SIZE: u32 = 100;
#[derive(Debug, FromRow)]
struct TaskRunRow {
id: i64,
task_id: String,
trigger_type: String,
status: String,
started_at: Option<DateTime<Utc>>,
finished_at: Option<DateTime<Utc>>,
duration_ms: Option<i64>,
exit_code: Option<i64>,
error_message: Option<String>,
created_at: DateTime<Utc>,
}
impl TryFrom<TaskRunRow> for TaskRunResponse {
type Error = ApiError;
fn try_from(row: TaskRunRow) -> Result<Self, Self::Error> {
Ok(Self {
id: row.id,
task_id: row.task_id,
trigger: parse_trigger(&row.trigger_type)?,
status: parse_status(&row.status)?,
started_at: row.started_at,
finished_at: row.finished_at,
duration_ms: row
.duration_ms
.map(u64::try_from)
.transpose()
.map_err(ApiError::internal)?,
exit_code: row
.exit_code
.map(i32::try_from)
.transpose()
.map_err(ApiError::internal)?,
error_message: row.error_message,
created_at: row.created_at,
})
}
}
pub async fn create_queued(
pool: &SqlitePool,
task_id: &str,
trigger: RunTrigger,
) -> ApiResult<TaskRunResponse> {
let now = Utc::now();
let result = sqlx::query(
"INSERT INTO task_runs (
task_id, trigger_type, status, created_at
) VALUES (?, ?, ?, ?)",
)
.bind(task_id)
.bind(trigger.as_str())
.bind(RunStatus::Queued.as_str())
.bind(now)
.execute(pool)
.await?;
Ok(TaskRunResponse {
id: result.last_insert_rowid(),
task_id: task_id.to_owned(),
trigger,
status: RunStatus::Queued,
started_at: None,
finished_at: None,
duration_ms: None,
exit_code: None,
error_message: None,
created_at: now,
})
}
pub async fn create_skipped(
pool: &SqlitePool,
task_id: &str,
trigger: RunTrigger,
message: &str,
) -> ApiResult<TaskRunResponse> {
let now = Utc::now();
let result = sqlx::query(
"INSERT INTO task_runs (
task_id, trigger_type, status, started_at, finished_at,
duration_ms, error_message, created_at
) VALUES (?, ?, ?, ?, ?, 0, ?, ?)",
)
.bind(task_id)
.bind(trigger.as_str())
.bind(RunStatus::Skipped.as_str())
.bind(now)
.bind(now)
.bind(message)
.bind(now)
.execute(pool)
.await?;
Ok(TaskRunResponse {
id: result.last_insert_rowid(),
task_id: task_id.to_owned(),
trigger,
status: RunStatus::Skipped,
started_at: Some(now),
finished_at: Some(now),
duration_ms: Some(0),
exit_code: None,
error_message: Some(message.to_owned()),
created_at: now,
})
}
/// 返回是否真的更新到了运行记录;记录被删除(例如任务被删)时返回 false。
pub async fn mark_running(
pool: &SqlitePool,
run_id: i64,
started_at: DateTime<Utc>,
) -> ApiResult<bool> {
let result = sqlx::query("UPDATE task_runs SET status = ?, started_at = ? WHERE id = ?")
.bind(RunStatus::Running.as_str())
.bind(started_at)
.bind(run_id)
.execute(pool)
.await?;
Ok(result.rows_affected() > 0)
}
/// 返回是否真的更新到了运行记录。
pub async fn complete(
pool: &SqlitePool,
run_id: i64,
status: RunStatus,
finished_at: DateTime<Utc>,
duration_ms: u64,
exit_code: Option<i32>,
error_message: Option<&str>,
) -> ApiResult<bool> {
let result = sqlx::query(
"UPDATE task_runs SET
status = ?, finished_at = ?, duration_ms = ?,
exit_code = ?, error_message = ?
WHERE id = ?",
)
.bind(status.as_str())
.bind(finished_at)
.bind(duration_ms as i64)
.bind(exit_code.map(i64::from))
.bind(error_message)
.bind(run_id)
.execute(pool)
.await?;
Ok(result.rows_affected() > 0)
}
/// 启动时收敛上一次进程遗留下来的非终态运行,避免界面永远显示"运行中"。
pub async fn fail_interrupted_runs(pool: &SqlitePool) -> ApiResult<u64> {
let result = sqlx::query(
"UPDATE task_runs SET status = ?, finished_at = ?, error_message = ?
WHERE status IN ('queued', 'running')",
)
.bind(RunStatus::Cancelled.as_str())
.bind(Utc::now())
.bind("TaskPilot 进程重启,运行被中断")
.execute(pool)
.await?;
Ok(result.rows_affected())
}
pub async fn get(pool: &SqlitePool, run_id: i64) -> ApiResult<Option<TaskRunResponse>> {
let row = sqlx::query_as::<_, TaskRunRow>(
"SELECT
id, task_id, trigger_type, status, started_at, finished_at,
duration_ms, exit_code, error_message, created_at
FROM task_runs
WHERE id = ?",
)
.bind(run_id)
.fetch_optional(pool)
.await?;
row.map(TaskRunResponse::try_from).transpose()
}
pub async fn list_for_task(
pool: &SqlitePool,
task_id: &str,
query: RunListQuery,
) -> ApiResult<TaskRunListResponse> {
let page = query.page.unwrap_or(1).max(1);
let page_size = query
.page_size
.unwrap_or(DEFAULT_PAGE_SIZE)
.clamp(1, MAX_PAGE_SIZE);
let offset = i64::from(page - 1) * i64::from(page_size);
let status = query.status.map(RunStatus::as_str);
let trigger = query.trigger.map(RunTrigger::as_str);
let total: i64 = sqlx::query_scalar(
"SELECT COUNT(*) FROM task_runs
WHERE task_id = ?
AND (? IS NULL OR status = ?)
AND (? IS NULL OR trigger_type = ?)",
)
.bind(task_id)
.bind(status)
.bind(status)
.bind(trigger)
.bind(trigger)
.fetch_one(pool)
.await?;
let rows = sqlx::query_as::<_, TaskRunRow>(
"SELECT
id, task_id, trigger_type, status, started_at, finished_at,
duration_ms, exit_code, error_message, created_at
FROM task_runs
WHERE task_id = ?
AND (? IS NULL OR status = ?)
AND (? IS NULL OR trigger_type = ?)
ORDER BY created_at DESC, id DESC
LIMIT ? OFFSET ?",
)
.bind(task_id)
.bind(status)
.bind(status)
.bind(trigger)
.bind(trigger)
.bind(i64::from(page_size))
.bind(offset)
.fetch_all(pool)
.await?;
let items = rows
.into_iter()
.map(TaskRunResponse::try_from)
.collect::<ApiResult<Vec<_>>>()?;
Ok(TaskRunListResponse {
items,
page,
page_size,
total: u64::try_from(total).map_err(ApiError::internal)?,
})
}
fn parse_trigger(value: &str) -> ApiResult<RunTrigger> {
match value {
"schedule" => Ok(RunTrigger::Schedule),
"manual" => Ok(RunTrigger::Manual),
"retry" => Ok(RunTrigger::Retry),
_ => Err(ApiError::internal(format!("unknown run trigger: {value}"))),
}
}
fn parse_status(value: &str) -> ApiResult<RunStatus> {
match value {
"queued" => Ok(RunStatus::Queued),
"running" => Ok(RunStatus::Running),
"succeeded" => Ok(RunStatus::Succeeded),
"failed" => Ok(RunStatus::Failed),
"timed_out" => Ok(RunStatus::TimedOut),
"cancelled" => Ok(RunStatus::Cancelled),
"skipped" => Ok(RunStatus::Skipped),
_ => Err(ApiError::internal(format!("unknown run status: {value}"))),
}
}