Replace log query implementation

The interface provided by Winston is unusably slow for querying large
logs.  Reimplement the log querying with both forwards and backwards
line reading and a simple file based index to skip files which do not
contain any lines matching the filter.
This commit is contained in:
Hornwitser committed 2021-12-19 02:06:59 +01:00
1 parent fd1f3329bd
commit 601ebba919
10 files changed
+497 -54

No files matched your search

+1 -2
View File
@@ -1,7 +1,6 @@
# Logs
logs
*.log
npm-debug.log*
*debug.log*
# Runtime data
pids
+322
View File
@@ -1,3 +1,9 @@
/**
* Utilities for logging on the Node.js side
*
* @author Hornwitser
* @module lib/logging_utils
*/
"use strict";
const events = require("events");
const fs = require("fs-extra");
@@ -9,6 +15,8 @@ const chalk = require("chalk");
const libLink = require("./link");
const libErrors = require("./errors");
const libFileOps = require("./file_ops");
const { levels, logger } = require("./logging");
const libStream = require("./stream");
@@ -59,6 +67,7 @@ function formatServerOutput(parsed) {
/**
* Formats winston log messages for a character terminal.
* @static
*/
class TerminalFormat {
constructor(options = {}) {
@@ -99,6 +108,7 @@ class TerminalFormat {
/**
* Sends logs over a lib/link connection.
* @static
*/
class LinkTransport extends Transport {
constructor(options) {
@@ -125,6 +135,315 @@ class LinkTransport extends Transport {
}
}
const logFileGlob = /^[a-z]+-(\d{4}-\d{2}-\d{2})\.log$/;
/**
* Filter object for logs
* @typedef {Object} module:lib/logging_utils~LogFilter
* @property {string} [max_level] -
* Maximum log level to include. Higher levels are more verbose.
* @property {boolean} [all] -
* Include log entries from master, all slaves and all instances.
* @property {boolean} [master] -
* Include log entries from the master server.
* @property {Array<number>} [slave_ids] -
* Include log entries for the given slaves and instances of those
* slaves by id.
* @property {Array<number>} [instance_ids] -
* Include log entries for the given instances by id.
*/
/**
* Keeps an index over a log directory to speed up queries to it
* @static
*/
class LogIndex {
constructor(logDirectory, serialized) {
this.logDirectory = logDirectory;
this.index = new Map();
if (serialized.version !== 0) {
return;
}
for (let [file, serializedEntry] of Object.entries(serialized.files)) {
this.index.set(file, {
levels: new Set(serializedEntry.levels),
master: serializedEntry.master,
slave_ids: new Set(serializedEntry.slave_ids),
instance_ids: new Set(serializedEntry.instance_ids),
});
}
}
serialize() {
let serialized = {
version: 0,
files: {},
};
for (let [file, entry] of this.index) {
serialized.files[file] = {
levels: [...entry.levels],
master: entry.master,
slave_ids: [...entry.slave_ids],
instance_ids: [...entry.instance_ids],
};
}
return serialized;
}
/**
* Load or create an empty index for the given directory
*
* Loads index.json from the given log directy expecting it to contain
* data about previously indexed files saved with {@link
* module:lib/logging_utils.LogIndex#save}.
*
* @param {string} logDirectory - Path to directory to load from.
* @returns {module:lib/logging_utils.LogIndex} loaded or created log index.
*/
static async load(logDirectory) {
try {
let content = await fs.readFile(path.join(logDirectory, "index.json"));
return new LogIndex(logDirectory, JSON.parse(content));
} catch (err) {
if (err.code !== "ENOENT") {
logger.warn(`Failed to load ${path.join(logDirectory, "index.json")}: ${err}`);
}
return new LogIndex(logDirectory, {});
}
}
/**
* Save index
*
* Saves corrently held index into index.json in the log directory it
* was loaded/created from.
*/
async save() {
await libFileOps.safeOutputFile(
path.join(this.logDirectory, "index.json"),
JSON.stringify(this.serialize(), null, 4)
);
}
async indexFile(file) {
let filePath = path.join(this.logDirectory, file);
let lineStream = new libStream.LineSplitter({ readableObjectMode: true });
let fileStream = fs.createReadStream(filePath);
fileStream.pipe(lineStream);
let entry = {
levels: new Set(),
master: false,
slave_ids: new Set(),
instance_ids: new Set(),
};
for await (let line of lineStream) {
let info;
try {
info = JSON.parse(line);
} catch (err) {
entry.levels.add("info");
entry.master = true;
continue;
}
if (info.level) {
entry.levels.add(info.level);
}
if (info.slave_id === undefined) {
entry.master = true;
} else if (info.instance_id === undefined) {
entry.slave_ids.add(info.slave_id);
}
if (info.instance_id !== undefined) {
entry.instance_ids.add(info.instance_id);
}
}
return entry;
}
/**
* Build index for files currently present in the log directory
*/
async buildIndex() {
let today = new Date(Date.now() - 300e3).toISOString().slice(0, 10);
for (let file of await fs.readdir(this.logDirectory)) {
let match = logFileGlob.exec(file);
if (!match || match[1] >= today || this.index.has(file)) {
continue;
}
let entry = await this.indexFile(file);
this.index.set(file, entry);
}
}
/**
* Check if a file includes entries for the given filter
*
* @param {string} file -
* Name of file in the log directory to check for.
* @param {module:lib/logging_utils~LogFilter} filter -
* Filter to check index if file contains entries for.
* @returns {boolean}
* true if the file may contain entries included by the filter.
*/
filterIncludesFile(file, { max_level, all, master, instance_ids, slave_ids }) {
let entry = this.index.get(file);
if (!entry) {
return true;
}
const hasOwnProperty = Object.prototype.hasOwnProperty;
if (max_level && hasOwnProperty.call(levels, max_level)) {
let fileLevels = [...entry.levels].map(
fileLevel => (hasOwnProperty.call(levels, fileLevel) ? levels[fileLevel] : levels.verbose)
);
if (fileLevels.every(level => level > levels[max_level])) {
return false;
}
}
if (all) {
return true;
}
if (master && entry.master) {
return true;
}
if (slave_ids && slave_ids.some(id => entry.slave_ids.has(id))) {
return true;
}
if (instance_ids && instance_ids.some(id => entry.instance_ids.has(id))) {
return true;
}
return false;
}
}
/**
* Create log filter by level and source.
*
* @param {module:lib/logging_utils~LogFilter} filter -
* Filter to filter log entries by.
* @returns {function(object): boolean}
* filter returning true for log entries that match it.
* @static
*/
function logFilter({ all, master, slave_ids, instance_ids, max_level }) {
return info => {
// Note: reversed to filter out undefined levels
if (max_level && !(levels[info.level] <= levels[max_level])) {
return false;
}
if (all) {
return true;
}
if (master && info.slave_id === undefined) {
return true;
}
if (
slave_ids
&& info.slave_id !== undefined
&& info.instance_id === undefined
&& slave_ids.includes(info.slave_id)
) {
return true;
}
if (instance_ids && info.instance_id !== undefined && instance_ids.includes(info.instance_id)) {
return true;
}
return false;
};
}
/**
* Filter object for querying logs
* @typedef {Object} module:lib/logging_utils~QueryLogFilter
* @property {number} [limit=100] - Maximum number of entries to return.
* @property {string} [order="asc"] -
* Return entries in ascending ("asc") date order or desceding ("desc")
* date order.
* @property {string} [max_level] -
* Maximum log level to include. Higher levels are more verbose.
* @property {boolean} [all] -
* Include log entries from master, all slaves and all instances.
* @property {boolean} [master] -
* Include log entries from the master server.
* @property {Array<number>} [slave_ids] -
* Include log entries for the given slaves and instances of those
* slaves by id.
* @property {Array<number>} [instance_ids] -
* Include log entries for the given instances by id.
*/
/**
* Query log directory
*
* @param {string} logDirectory - path to directory with logs.
* @param {module:lib/logging_utils~QueryLogFilter} filter -
* Filter to limit logs by.
* @param {module:lib/logging_utils.LogIndex=} index -
* Index to speed up query with.
* @returns {Array<Object>} log entries matching the filter
*/
async function queryLog(logDirectory, filter, index) {
let files = (await fs.readdir(logDirectory)).filter(entry => logFileGlob.test(entry));
filter = {
limit: 100,
order: "asc",
...filter,
};
files.sort();
if (filter.order === "desc") {
files.reverse();
}
const includeEntry = logFilter(filter);
let log = [];
for (let file of files) {
if (index && !index.filterIncludesFile(file, filter)) {
continue;
}
let filePath = path.join(logDirectory, file);
let fileStream;
let lineStream;
if (filter.order === "asc") {
lineStream = new libStream.LineSplitter({ readableObjectMode: true });
fileStream = fs.createReadStream(filePath);
} else {
lineStream = new libStream.ReverseLineSplitter({ readableObjectMode: true });
fileStream = await libStream.createReverseReadStream(filePath);
}
lineStream.on("data", line => {
let info;
try {
info = JSON.parse(line);
} catch (err) {
info = { level: "info", message: line.toString("utf8") };
}
if (includeEntry(info)) {
log.push(info);
if (log.length >= filter.limit) {
lineStream.destroy();
fileStream.destroy();
}
}
});
fileStream.pipe(lineStream);
await events.once(lineStream, "close");
if (log.length >= filter.limit) {
break;
}
}
return log;
}
async function migrateLogs(log, directory, pattern) {
const dateRegExp = /^\d{4}-\d{2}-\d{2}/;
let lineStream = new libStream.LineSplitter({ readableObjectMode: true });
@@ -185,6 +504,9 @@ async function migrateLogs(log, directory, pattern) {
module.exports = {
TerminalFormat,
LinkTransport,
LogIndex,
logFilter,
queryLog,
migrateLogs,
// for testing only
+9 -52
View File
@@ -1,12 +1,10 @@
"use strict";
const jwt = require("jsonwebtoken");
const util = require("util");
const winston = require("winston");
const libConfig = require("@clusterio/lib/config");
const libErrors = require("@clusterio/lib/errors");
const libLink = require("@clusterio/lib/link");
const { levels, logger } = require("@clusterio/lib/logging");
const { logger } = require("@clusterio/lib/logging");
const libLoggingUtils = require("@clusterio/lib/logging_utils");
const libPlugin = require("@clusterio/lib/plugin");
const libPrometheus = require("@clusterio/lib/prometheus");
@@ -402,29 +400,6 @@ class ControlConnection extends BaseConnection {
this.updateLogSubscriptions();
}
static logFilter({ all, master, slave_ids, instance_ids, max_level }) {
return info => {
// Note: reversed to filter out undefined levels
if (max_level && !(levels[info.level] <= levels[max_level])) {
return false;
}
if (all) {
return true;
}
if (master && info.slave_id === undefined) {
return true;
}
if (info.slave_id && slave_ids.includes(info.slave_id)) {
return true;
}
if (info.instance_id && instance_ids.includes(info.instance_id)) {
return true;
}
return false;
};
}
updateLogSubscriptions() {
let { all, master, slave_ids, instance_ids } = this.logSubscriptions;
if (all || master || slave_ids.length || instance_ids.length) {
@@ -432,7 +407,7 @@ class ControlConnection extends BaseConnection {
this.logTransport = new libLoggingUtils.LinkTransport({ link: this });
this._master.clusterLogger.add(this.logTransport);
}
this.logTransport.filter = this.constructor.logFilter(this.logSubscriptions);
this.logTransport.filter = libLoggingUtils.logFilter(this.logSubscriptions);
} else if (this.logTransport) {
this._master.clusterLogger.remove(this.logTransport);
@@ -441,34 +416,16 @@ class ControlConnection extends BaseConnection {
}
async queryLogRequestHandler(message) {
let transport = new winston.transports.File({ filename: "cluster.log" });
let query = util.promisify((options, callback) => transport.query(options, callback));
// Available options and defaults inferred from reading the source
// rows: number = limit || 10
// limit: number // alias of rows.
// start: number = 0
// until: Date = now
// from: Date = until - 24h
// order: "asc"|"desc" = "desc"
// fields: Array<string>
// level: string
// This interface is junk:
// - The level option filters by exact match
// - No way to add your own filter
// - No way to stream results
// - The log file is read starting from the beginning
let setDuration = lastQueryLogTime.startTimer();
let log = await query({
order: message.data.order,
limit: Infinity,
from: new Date(0),
});
let { all, master, slave_ids, instance_ids } = message.data;
log = log.filter(this.constructor.logFilter(message.data));
if (message.data.limit) {
log = log.slice(0, message.data.limit);
let log;
if (!all && master && !slave_ids.length && !instance_ids.length) {
log = await this._master.queryMasterLog(message.data);
} else {
log = await this._master.queryClusterLog(message.data);
}
setDuration();
return { log };
}
+40
View File
@@ -12,6 +12,7 @@ const libConfig = require("@clusterio/lib/config");
const libErrors = require("@clusterio/lib/errors");
const libFileOps = require("@clusterio/lib/file_ops");
const libLink = require("@clusterio/lib/link");
const libLoggingUtils = require("@clusterio/lib/logging_utils");
const libPlugin = require("@clusterio/lib/plugin");
const libPluginLoader = require("@clusterio/lib/plugin_loader");
const libPrometheus = require("@clusterio/lib/prometheus");
@@ -117,6 +118,14 @@ class Master {
}
async _startInternal(args) {
this.logDirectory = args.logDirectory;
this.clusterLogIndex = await libLoggingUtils.LogIndex.load(path.join(this.logDirectory, "cluster"));
this.clusterLogBuildInterval = setInterval(() => {
this.clusterLogIndex.buildIndex().catch(
err => logger.error(`Error building cluster log index:\n${err.stack}`)
);
}, 600e3);
// Start webpack development server if enabled
if (args.dev || args.devPlugin) {
logger.warn("Webpack development mode enabled");
@@ -240,6 +249,10 @@ class Master {
// This function should never throw.
this._state = "stopping";
logger.info("Stopping master");
clearInterval(this.clusterLogBuildInterval);
if (this.clusterLogIndex) {
await this.clusterLogIndex.save();
}
logger.info("Saving config");
await libFileOps.safeOutputFile(this.configPath, JSON.stringify(this.config.serialize(), null, 4));
@@ -333,6 +346,33 @@ class Master {
await libFileOps.safeOutputFile(filePath, JSON.stringify(serialized, null, 4));
}
/**
* Query master log
*
* @param {module:lib/logging_utils~QueryLogFilter} filter -
* Filter to limit entries with. Note that only the master log can
* be queried from this function.
* @returns {Array<Object>} log entries matching the filter
*/
async queryMasterLog(filter) {
return await libLoggingUtils.queryLog(
path.join(this.logDirectory, "master"), filter,
);
}
/**
* Query cluster log
*
* @param {module:lib/logging_utils~QueryLogFilter} filter -
* Filter to limit entries with.
* @returns {Array<Object>} log entries matching the filter
*/
async queryClusterLog(filter) {
return await libLoggingUtils.queryLog(
path.join(this.logDirectory, "cluster"), filter, this.clusterLogIndex,
);
}
static addAppRoutes(app, pluginInfos) {
app.use((req, res, next) => {
hitCounter.inc();
+1
View File
@@ -0,0 +1 @@
This file should be excluded by queries and the index
+9
View File
@@ -0,0 +1,9 @@
{"timestamp":"2021-12-10T10:00:00.000Z","level":"verbose","message":"Master start"}
{"timestamp":"2021-12-10T10:00:01.000Z","level":"verbose","message":"Loading"}
{"timestamp":"2021-12-10T10:00:03.000Z","level":"info","message":"Ready"}
{"timestamp":"2021-12-10T10:00:10.000Z","slave_id":1,"level":"verbose","message":"Slave connected"}
{"timestamp":"2021-12-10T10:00:11.000Z","slave_id":1,"level":"verbose","message":"Found instance 10"}
{"timestamp":"2021-12-10T10:00:13.000Z","slave_id":1,"instance_id":10,"level":"server","message":"Starting up"}
{"timestamp":"2021-12-10T10:00:14.000Z","slave_id":1,"instance_id":10,"level":"server","message":"Loading map"}
{"timestamp":"2021-12-10T10:00:15.000Z","slave_id":1,"instance_id":10,"level":"server","message":"Game ready"}
{"timestamp":"2021-12-10T10:01:00.000Z","slave_id":1,"instance_id":10,"level":"error","message":"Unexpected crash"}
+3
View File
@@ -0,0 +1,3 @@
{"timestamp":"2021-12-11T10:00:00.000Z","level":"verbose","message":"Master start"}
{"timestamp":"2021-12-11T10:00:01.000Z","level":"verbose","message":"Loading"}
{"timestamp":"2021-12-11T10:00:03.000Z","level":"fatal","message":"An error occured"}
+1
View File
@@ -0,0 +1 @@
{"timestamp":"2021-12-12T10:00:00.000Z","leve
View File
Whitespace-only changes.
+111
View File
@@ -1,8 +1,13 @@
"use strict";
const assert = require("assert").strict;
const fs = require("fs-extra");
const path = require("path");
const util = require("util");
const { testLines } = require("./factorio/lines");
const libLoggingUtils = require("@clusterio/lib/logging_utils");
const libStream = require("@clusterio/lib/stream");
describe("lib/logging_utils.js", function() {
@@ -16,4 +21,110 @@ describe("lib/logging_utils.js", function() {
}
});
});
let logDir = path.join("test", "file", "log");
describe("class LogIndex", function() {
it("should create a new empty index if index.json does not exist", async function() {
let logIndex = await libLoggingUtils.LogIndex.load(logDir);
assert.deepEqual(logIndex.index, new Map());
});
it("should build index of only past log files", async function() {
let logIndex = await libLoggingUtils.LogIndex.load(logDir);
await logIndex.buildIndex();
assert.deepEqual(
new Set(logIndex.index.keys()),
new Set(["test-2021-12-10.log", "test-2021-12-11.log", "test-2021-12-12.log"])
);
});
it("should roundtrip index on save load cycle", async function() {
let logIndex = await libLoggingUtils.LogIndex.load(logDir);
await logIndex.buildIndex();
let tempDir = path.join("temp", "test", "log");
logIndex.logDirectory = tempDir;
await logIndex.save();
let loadedIndex = await libLoggingUtils.LogIndex.load(tempDir);
assert.deepEqual(
loadedIndex,
logIndex
);
});
it("should handle loading broken index", async function() {
let tempDir = path.join("temp", "test", "log");
await fs.outputFile(path.join(tempDir, "index.json"), "Broken JSON");
let logIndex = await libLoggingUtils.LogIndex.load(tempDir);
assert.deepEqual(logIndex.index, new Map());
});
});
let allLines = [];
let logIndex = null;
before(async function() {
for (let file of (await fs.readdir(logDir)).sort()) {
if (file === "excluded.txt") {
continue;
}
let fileStream = fs.createReadStream(path.join(logDir, file));
let lineStream = new libStream.LineSplitter({ readableObjectMode: true });
fileStream.pipe(lineStream);
for await (let line of lineStream) {
try {
allLines.push(JSON.parse(line));
} catch (err) {
allLines.push({ "level": "info", "message": line.toString() });
}
}
}
logIndex = await libLoggingUtils.LogIndex.load(logDir);
await logIndex.buildIndex();
});
describe("queryLog", function() {
for (let index of [() => undefined, () => logIndex]) {
describe(index() === undefined ? "without index" : "with index", function() {
it("returns nothing with an empty filter", async function() {
let log = await libLoggingUtils.queryLog(logDir, {}, index());
assert.deepEqual(log, []);
});
it("should return the whole log when querying all", async function() {
let log = await libLoggingUtils.queryLog(logDir, { all: true }, index());
assert.deepEqual(log, allLines);
});
it("should return the whole log in reverse when querying all desc", async function() {
let log = await libLoggingUtils.queryLog(logDir, { all: true, order: "desc" }, index());
assert.deepEqual(log, [...allLines].reverse());
});
it("should limit by limit", async function() {
let log = await libLoggingUtils.queryLog(logDir, { limit: 2, all: true }, index());
assert.deepEqual(log, allLines.slice(0, 2));
});
it("should filter by max_level", async function() {
let log = await libLoggingUtils.queryLog(logDir, { max_level: "info", all: true }, index());
assert.deepEqual(log, allLines.filter(
info => ["fatal", "error", "warn", "audit", "info"].includes(info.level)
));
log = await libLoggingUtils.queryLog(logDir, { max_level: "fatal", all: true }, index());
assert.deepEqual(log, allLines.filter(
info => ["fatal"].includes(info.level)
));
});
it("should filter by master", async function() {
let log = await libLoggingUtils.queryLog(logDir, { master: true }, index());
assert.deepEqual(log, allLines.filter(
info => info.slave_id === undefined && info.instance_id === undefined
));
});
it("should filter by slave_ids", async function() {
let log = await libLoggingUtils.queryLog(logDir, { slave_ids: [1, 2] }, index());
assert.deepEqual(log, allLines.filter(
info => [1, 2].includes(info.slave_id) && info.instance_id === undefined
));
});
it("should filter by instance_ids", async function() {
let log = await libLoggingUtils.queryLog(logDir, { instance_ids: [10, 11] }, index());
assert.deepEqual(log, allLines.filter(info => [10, 11].includes(info.instance_id)));
});
});
}
});
});