Merge 5e55215c9c into c01494ec33

# Conflicts:
#	server/database.js
#	server/jobs/clear-old-data.js

extra/reset-migrate-aggregate-table-state.js

@@ -0,0 +1,24 @@
+const { R } = require("redbean-node");
+const Database = require("../server/database");
+const args = require("args-parser")(process.argv);
+const { Settings } = require("../server/settings");
+
+const main = async () => {
+    console.log("Connecting the database");
+    Database.initDataDir(args);
+    await Database.connect(false, false, true);
+
+    console.log("Deleting all data from aggregate tables");
+    await R.exec("DELETE FROM stat_minutely");
+    await R.exec("DELETE FROM stat_hourly");
+    await R.exec("DELETE FROM stat_daily");
+
+    console.log("Resetting the aggregate table state");
+    await Settings.set("migrateAggregateTableState", "");
+
+    await Database.close();
+    console.log("Done");
+};
+
+main();
+

package.json

@@ -68,7 +68,8 @@
         "sort-contributors": "node extra/sort-contributors.js",
         "quick-run-nightly": "docker run --rm --env NODE_ENV=development -p 3001:3001 louislam/uptime-kuma:nightly2",
         "start-dev-container": "cd docker && docker-compose -f docker-compose-dev.yml up --force-recreate",
-        "rebase-pr-to-1.23.X": "node extra/rebase-pr.js 1.23.X"
+        "rebase-pr-to-1.23.X": "node extra/rebase-pr.js 1.23.X",
+        "reset-migrate-aggregate-table-state": "node extra/reset-migrate-aggregate-table-state.js"
     },
     "dependencies": {
         "@grpc/grpc-js": "~1.8.22",

server/database.js

@@ -6,6 +6,9 @@ const knex = require("knex");
 const path = require("path");
 const { EmbeddedMariaDB } = require("./embedded-mariadb");
 const mysql = require("mysql2/promise");
+const { Settings } = require("./settings");
+const { UptimeCalculator } = require("./uptime-calculator");
+const dayjs = require("dayjs");
 
 /**
  * Database & App Data Folder
@@ -391,9 +394,23 @@ class Database {
         // https://knexjs.org/guide/migrations.html
         // https://gist.github.com/NigelEarle/70db130cc040cc2868555b29a0278261
         try {
+            // Disable foreign key check for SQLite
+            // Known issue of knex: https://github.com/drizzle-team/drizzle-orm/issues/1813
+            if (Database.dbConfig.type === "sqlite") {
+                await R.exec("PRAGMA foreign_keys = OFF");
+            }
+
             await R.knex.migrate.latest({
                 directory: Database.knexMigrationsPath,
             });
+
+            // Enable foreign key check for SQLite
+            if (Database.dbConfig.type === "sqlite") {
+                await R.exec("PRAGMA foreign_keys = ON");
+            }
+
+            await this.migrateAggregateTable();
+
         } catch (e) {
             // Allow missing patch files for downgrade or testing pr.
             if (e.message.includes("the following files are missing:")) {
@@ -711,6 +728,119 @@ class Database {
         }
     }
 
+    /**
+     * Migrate the old data in the heartbeat table to the new format (stat_daily, stat_hourly, stat_minutely)
+     * It should be run once while upgrading V1 to V2
+     *
+     * Normally, it should be in transaction, but UptimeCalculator wasn't designed to be in transaction before that.
+     * I don't want to heavily modify the UptimeCalculator, so it is not in transaction.
+     * Run `npm run reset-migrate-aggregate-table-state` to reset, in case the migration is interrupted.
+     * @returns {Promise<void>}
+     */
+    static async migrateAggregateTable() {
+        log.debug("db", "Enter Migrate Aggregate Table function");
+
+        // Add a setting for 2.0.0-dev users to skip this migration
+        if (process.env.SET_MIGRATE_AGGREGATE_TABLE_TO_TRUE === "1") {
+            log.warn("db", "SET_MIGRATE_AGGREGATE_TABLE_TO_TRUE is set to 1, skipping aggregate table migration forever (for 2.0.0-dev users)");
+            await Settings.set("migrateAggregateTableState", "migrated");
+        }
+
+        let migrateState = await Settings.get("migrateAggregateTableState");
+
+        // Skip if already migrated
+        // If it is migrating, it possibly means the migration was interrupted, or the migration is in progress
+        if (migrateState === "migrated") {
+            log.debug("db", "Migrated aggregate table already, skip");
+            return;
+        } else if (migrateState === "migrating") {
+            log.warn("db", "Aggregate table migration is already in progress, or it was interrupted");
+            throw new Error("Aggregate table migration is already in progress");
+        }
+
+        await Settings.set("migrateAggregateTableState", "migrating");
+
+        log.info("db", "Migrating Aggregate Table");
+
+        log.info("db", "Getting list of unique monitors");
+
+        // Get a list of unique monitors from the heartbeat table, using raw sql
+        let monitors = await R.getAll(`
+            SELECT DISTINCT monitor_id
+            FROM heartbeat
+            ORDER BY monitor_id ASC
+        `);
+
+        // Stop if stat_* tables are not empty
+        for (let table of [ "stat_minutely", "stat_hourly", "stat_daily" ]) {
+            let countResult = await R.getRow(`SELECT COUNT(*) AS count FROM ${table}`);
+            let count = countResult.count;
+            if (count > 0) {
+                log.warn("db", `Aggregate table ${table} is not empty, migration will not be started (Maybe you were using 2.0.0-dev?)`);
+                return;
+            }
+        }
+
+        let progressPercent = 0;
+        let part = 100 / monitors.length;
+        let i = 1;
+        for (let monitor of monitors) {
+            // Get a list of unique dates from the heartbeat table, using raw sql
+            let dates = await R.getAll(`
+                SELECT DISTINCT DATE(time) AS date
+                FROM heartbeat
+                WHERE monitor_id = ?
+                ORDER BY date ASC
+            `, [
+                monitor.monitor_id
+            ]);
+
+            for (let date of dates) {
+                // New Uptime Calculator
+                let calculator = new UptimeCalculator();
+                calculator.monitorID = monitor.monitor_id;
+                calculator.setMigrationMode(true);
+
+                // Get all the heartbeats for this monitor and date
+                let heartbeats = await R.getAll(`
+                    SELECT status, ping, time
+                    FROM heartbeat
+                    WHERE monitor_id = ?
+                    AND DATE(time) = ?
+                    ORDER BY time ASC
+                `, [ monitor.monitor_id, date.date ]);
+
+                if (heartbeats.length > 0) {
+                    log.info("db", `[DON'T STOP] Migrating monitor data ${monitor.monitor_id} - ${date.date} [${progressPercent.toFixed(2)}%][${i}/${monitors.length}]`);
+                }
+
+                for (let heartbeat of heartbeats) {
+                    await calculator.update(heartbeat.status, parseFloat(heartbeat.ping), dayjs(heartbeat.time));
+                }
+
+                progressPercent += (Math.round(part / dates.length * 100) / 100);
+
+                // Lazy to fix the floating point issue, it is acceptable since it is just a progress bar
+                if (progressPercent > 100) {
+                    progressPercent = 100;
+                }
+            }
+
+            i++;
+        }
+
+        // TODO: Remove all non-important heartbeats from heartbeat table
+        log.info("db", "[DON'T STOP] Deleting all data from heartbeat table");
+
+        await Settings.set("migrateAggregateTableState", "migrated");
+
+        if (monitors.length > 0) {
+            log.info("db", "Aggregate Table Migration Completed");
+        } else {
+            log.info("db", "No data to migrate");
+        }
+    }
+
 }
 
 module.exports = Database;

server/jobs/clear-old-data.js

@@ -2,6 +2,7 @@ const { R } = require("redbean-node");
 const { log } = require("../../src/util");
 const { setSetting, setting } = require("../util-server");
 const Database = require("../database");
+const { Settings } = require("../settings");
 
 const DEFAULT_KEEP_PERIOD = 180;
 
@@ -11,11 +12,28 @@ const DEFAULT_KEEP_PERIOD = 180;
  */
 
 const clearOldData = async () => {
+
+    // TODO: Temporary disable for testing
+    return;
+
+    /*
+    * TODO:
+    * Since we have aggregated table now, we don't need so much data in heartbeat table.
+    * But we still need to keep the important rows, because they contain the message.
+    *
+    * In the heartbeat table:
+    *    - important rows: keep according to the setting (keepDataPeriodDays) (default 180 days)
+    *    - not important rows: keep 2 days
+    *
+    * stat_* tables:
+    *   - keep according to the setting (keepDataPeriodDays) (default 180 days)
+    */
+
     let period = await setting("keepDataPeriodDays");
 
     // Set Default Period
     if (period == null) {
-        await setSetting("keepDataPeriodDays", DEFAULT_KEEP_PERIOD, "general");
+        await Settings.set("keepDataPeriodDays", DEFAULT_KEEP_PERIOD, "general");
         period = DEFAULT_KEEP_PERIOD;
     }
 

server/server.js

@@ -1604,18 +1604,20 @@ let needSetup = false;
 
     await server.start();
 
-    server.httpServer.listen(port, hostname, () => {
+    server.httpServer.listen(port, hostname, async () => {
         if (hostname) {
             log.info("server", `Listening on ${hostname}:${port}`);
         } else {
             log.info("server", `Listening on ${port}`);
         }
-        startMonitors();
+        await startMonitors();
+
+        // Put this here. Start background jobs after the db and server is ready to prevent clear up during db migration.
+        await initBackgroundJobs();
+
         checkVersion.startInterval();
     });
 
-    await initBackgroundJobs();
-
     // Start cloudflared at the end if configured
     await cloudflaredAutoStart(cloudflaredToken);
 
@@ -1809,7 +1811,11 @@ async function startMonitors() {
     }
 
     for (let monitor of list) {
+        try {
             await monitor.start(io);
+        } catch (e) {
+            log.error("monitor", e);
+        }
         // Give some delays, so all monitors won't make request at the same moment when just start the server.
         await sleep(getRandomInt(300, 1000));
     }

server/uptime-calculator.js

@@ -12,7 +12,6 @@ class UptimeCalculator {
      * @private
      * @type {{string:UptimeCalculator}}
      */
-
     static list = {};
 
     /**
@@ -55,6 +54,15 @@ class UptimeCalculator {
     lastHourlyStatBean = null;
     lastMinutelyStatBean = null;
 
+    /**
+     * For migration purposes.
+     * @type {boolean}
+     */
+    migrationMode = false;
+
+    statMinutelyKeepHour = 24;
+    statHourlyKeepDay = 30;
+
     /**
      * Get the uptime calculator for a monitor
      * Initializes and returns the monitor if it does not exist
@@ -189,11 +197,14 @@ class UptimeCalculator {
     /**
      * @param {number} status status
      * @param {number} ping Ping
+     * @param {dayjs.Dayjs} date Date (Only for migration)
      * @returns {dayjs.Dayjs} date
      * @throws {Error} Invalid status
      */
-    async update(status, ping = 0) {
-        let date = this.getCurrentDate();
+    async update(status, ping = 0, date) {
+        if (!date) {
+            date = this.getCurrentDate();
+        }
 
         let flatStatus = this.flatStatus(status);
 
@@ -297,6 +308,11 @@ class UptimeCalculator {
         }
         await R.store(dailyStatBean);
 
+        let currentDate = this.getCurrentDate();
+
+        // For migration mode, we don't need to store old hourly and minutely data, but we need 30-day's hourly data
+        // Run anyway for non-migration mode
+        if (!this.migrationMode || date.isAfter(currentDate.subtract(this.statHourlyKeepDay, "day"))) {
             let hourlyStatBean = await this.getHourlyStatBean(hourlyKey);
             hourlyStatBean.up = hourlyData.up;
             hourlyStatBean.down = hourlyData.down;
@@ -311,7 +327,11 @@ class UptimeCalculator {
                 }
             }
             await R.store(hourlyStatBean);
+        }
 
+        // For migration mode, we don't need to store old hourly and minutely data, but we need 24-hour's minutely data
+        // Run anyway for non-migration mode
+        if (!this.migrationMode || date.isAfter(currentDate.subtract(this.statMinutelyKeepHour, "hour"))) {
             let minutelyStatBean = await this.getMinutelyStatBean(divisionKey);
             minutelyStatBean.up = minutelyData.up;
             minutelyStatBean.down = minutelyData.down;
@@ -326,18 +346,23 @@ class UptimeCalculator {
                 }
             }
             await R.store(minutelyStatBean);
+        }
 
+        // No need to remove old data in migration mode
+        if (!this.migrationMode) {
             // Remove the old data
+            // TODO: Improvement: Convert it to a job?
             log.debug("uptime-calc", "Remove old data");
             await R.exec("DELETE FROM stat_minutely WHERE monitor_id = ? AND timestamp < ?", [
                 this.monitorID,
-            this.getMinutelyKey(date.subtract(24, "hour")),
+                this.getMinutelyKey(currentDate.subtract(this.statMinutelyKeepHour, "hour")),
             ]);
 
             await R.exec("DELETE FROM stat_hourly WHERE monitor_id = ? AND timestamp < ?", [
                 this.monitorID,
-            this.getHourlyKey(date.subtract(30, "day")),
+                this.getHourlyKey(currentDate.subtract(this.statHourlyKeepDay, "day")),
             ]);
+        }
 
         return date;
     }
@@ -812,6 +837,14 @@ class UptimeCalculator {
         return dayjs.utc();
     }
 
+    /**
+     * For migration purposes.
+     * @param {boolean} value Migration mode on/off
+     * @returns {void}
+     */
+    setMigrationMode(value) {
+        this.migrationMode = value;
+    }
 }
 
 class UptimeDataResult {