feat(backend): add JOB_RETENTION_SECS to delete completed jobs completed after expiration period

README.md

@@ -338,6 +338,7 @@ it being synced automatically everyday.
 | GLOBAL_CACHE_INTERVAL     | 10\*60                                     | (Enterprise Edition only) Interval in seconds in between bucket sync of the cache. This interval \* 2 is the time at which you're guaranteed all the worker's caches are synced together.          | Worker                |
 | WORKER_TAGS               | 'deno,go,python3,bash,flow,hub,dependency' | The worker groups assigned to that workers                                                                                                                                                         | Worker                |
 | CUSTOM_TAGS               | None                                       | The custom tags assignable to scripts.                                                                                                                                                             | Server                |
+| JOB_RETENTION_SECS        | 60*60*24\*60 //60 days                     | The time in seconds after which jobs get deleted. Set to 0 or -1 to never delete                                                                                                                   | Server                |
 
 ## Run a local dev setup
 

backend/sqlx-data.json

@@ -1945,6 +1945,26 @@
     },
     "query": "INSERT INTO pip_resolution_cache (hash, lockfile, expiration) VALUES ($1, $2, now() + ('3 days')::interval) ON CONFLICT (hash) DO UPDATE SET lockfile = $2"
   },
+  "502781c4e2fc692db7a66b850450a1934b2403a5cff0421fe9609f9b99d8ee95": {
+    "describe": {
+      "columns": [
+        {
+          "name": "id",
+          "ordinal": 0,
+          "type_info": "Uuid"
+        }
+      ],
+      "nullable": [
+        false
+      ],
+      "parameters": {
+        "Left": [
+          "Int4"
+        ]
+      }
+    },
+    "query": "DELETE FROM completed_job WHERE started_at + ((duration_ms/1000 + $1) || ' s')::interval <= now() RETURNING id"
+  },
   "5061c0d054bf4f028e7fe51a8f9389024c6ae4492755cadac0f7167e5300bda0": {
     "describe": {
       "columns": [],

backend/src/main.rs

@@ -16,6 +16,7 @@ use monitor::handle_zombie_jobs_periodically;
 use sqlx::{Pool, Postgres};
 use tokio::{
     fs::{metadata, DirBuilder},
+    join,
     sync::RwLock,
 };
 use windmill_common::{utils::rd_string, METRICS_ADDR};
@@ -164,6 +165,7 @@ Windmill Community Edition {GIT_VERSION}
         "GLOBAL_CACHE_INTERVAL",
         "WORKER_TAGS",
         "CUSTOM_TAGS",
+        "JOB_RETENTION_SECS",
     ]);
 
     if server_mode || num_workers > 0 {
@@ -246,9 +248,11 @@ pub fn monitor_db<R: rsmq_async::RsmqConnection + Send + Sync + Clone + 'static>
     let rx2 = rx.resubscribe();
     let base_internal_url = base_internal_url.to_string();
     tokio::spawn(async move {
-        handle_zombie_jobs_periodically(&db1, rx, &base_internal_url, rsmq).await
+        join!(
+            handle_zombie_jobs_periodically(&db1, rx, &base_internal_url, rsmq),
+            windmill_api::delete_expired_items_perdiodically(&db2, rx2)
+        );
     });
-    tokio::spawn(async move { windmill_api::delete_expired_items_perdiodically(&db2, rx2).await });
 }
 
 pub async fn run_workers<R: rsmq_async::RsmqConnection + Send + Sync + Clone + 'static>(

backend/windmill-worker/src/worker.rs

@@ -144,7 +144,7 @@ pub struct Metrics {
 }
 
 
-pub const DEFAULT_TIMEOUT: u16 = 300;
+pub const DEFAULT_TIMEOUT: u64 = 300;
 pub const DEFAULT_SLEEP_QUEUE: u64 = 50;
 
 lazy_static::lazy_static! {
@@ -213,12 +213,12 @@ lazy_static::lazy_static! {
         "Total number of seconds since the worker has started"
     );
 
-    static ref TIMEOUT: u16 = std::env::var("TIMEOUT")
+    static ref TIMEOUT: u64 = std::env::var("TIMEOUT")
         .ok()
-        .and_then(|x| x.parse::<u16>().ok())
-        .unwrap_or(DEFAULT_TIMEOUT as u16);
+        .and_then(|x| x.parse::<u64>().ok())
+        .unwrap_or(DEFAULT_TIMEOUT);
 
-    static ref TIMEOUT_DURATION: Duration = Duration::from_secs(*TIMEOUT as u64);
+    static ref TIMEOUT_DURATION: Duration = Duration::from_secs(*TIMEOUT);
 
     pub static ref SESSION_TOKEN_EXPIRY: i32 = (*TIMEOUT as i32) * 2;