@logdna/tail-file

What is @logdna/tail-file?

@logdna/tail-file is an npm package designed to efficiently tail files, meaning it can read and monitor the end of a file for new data. This is particularly useful for log files where new entries are continuously appended. The package provides a simple API to start, stop, and manage the tailing process, making it suitable for real-time log monitoring and processing.

What are @logdna/tail-file's main functionalities?

Basic Tailing

This feature allows you to start tailing a file and listen for new lines being added to the file. The 'line' event is emitted whenever a new line is appended to the file.

const { TailFile } = require('@logdna/tail-file');

const tail = new TailFile('/path/to/your/file.log');

(async () => {
  try {
    await tail.start();
    tail.on('line', (line) => {
      console.log('New line:', line);
    });
  } catch (err) {
    console.error('Error starting tail:', err);
  }
})();

Handling Errors

This feature demonstrates how to handle errors that may occur during the tailing process. The 'error' event is emitted whenever an error is encountered.

const { TailFile } = require('@logdna/tail-file');

const tail = new TailFile('/path/to/your/file.log');

tail.on('error', (err) => {
  console.error('Error:', err);
});

(async () => {
  try {
    await tail.start();
  } catch (err) {
    console.error('Error starting tail:', err);
  }
})();

Stopping the Tail

This feature shows how to stop the tailing process after a certain period. The 'quit' method is used to stop tailing the file.

const { TailFile } = require('@logdna/tail-file');

const tail = new TailFile('/path/to/your/file.log');

(async () => {
  try {
    await tail.start();
    // Stop tailing after 10 seconds
    setTimeout(async () => {
      await tail.quit();
      console.log('Stopped tailing');
    }, 10000);
  } catch (err) {
    console.error('Error starting tail:', err);
  }
})();

Other packages similar to @logdna/tail-file

tail

The 'tail' package is a simple library for tailing a file in Node.js. It provides basic functionality to read new lines appended to a file. Compared to @logdna/tail-file, it offers fewer features and less flexibility in handling errors and managing the tailing process.

tail-stream

The 'tail-stream' package is another alternative for tailing files in Node.js. It provides a stream interface for reading new lines from a file. While it is more flexible than the 'tail' package, it still lacks some of the advanced features and error handling capabilities provided by @logdna/tail-file.

chokidar

The 'chokidar' package is a file watcher library that can be used to monitor changes in files and directories. While it is not specifically designed for tailing files, it can be used to achieve similar functionality by watching for changes and reading new data. It offers a broader range of features for file system monitoring compared to @logdna/tail-file.

README

TailFile

At LogDNA, consuming log files and making them searchable is what we do! It all starts with the ability to efficiently watch log files on a local host and send new lines up to the LogDNA service. This Node.js class provides functionality like Unix's tail -f command, and we use it in our agents to get the job done. Of course, anything needing tail functionality in Node.js could also benefit from using this.

---

• Features
• Installation
• Usage
  • Example consuming data events
  • Example using pipe
  • Example using readline
  • Example for Clean Shutdown
• Events
  • Event: 'flush'
  • Event: 'renamed'
  • Event: 'retry'
  • Event: 'tail_error'
  • Event: 'truncated'
  • Event: (Any Readable event)
• API
  • Constructor: new TailLog(filename[, options])
  • tail.start()
  • tail.quit()
• Program Flow
• How Log Rolling is Handled
• Backpressure Pauses Polling
  • Log Rolling During Backpressure
• Contributing

Features

• Zero dependencies! It's lightweight and uses 100% Node.js core modules.
• It implements a Readable stream, which is efficient and flexible in terms of being able to pipe to other streams or consume via events.
• Stream backpressure is properly respected, so at no time is data pushed through the stream unless it is requested.
• It handles log rolling. Renaming files is handled gracefully without losing lines written to the "old" file, no matter what the poll interval is.
• It handles file truncation, continuing to tail the file despite being truncated.

Installation

npm install @logdna/tail-file

Supported Versions of Node.js

This package works with the following supported versions. Other versions might work, but the ones listed here are officially tested. If backward-compatibility is lost over time, a major version will be released supporting the latest version, and deprecating incompatible versions. Users are then encouraged to try previous versions of this package for those deprecated versions of Node.js.

Node.js versions: 18, 20

Usage

Instantiate an instance by passing the full path of a file to tail. This will return a stream that can be piped to other streams or consumed via data events. To begin the tailing, call the start method.

Example using data events

const TailFile = require('@logdna/tail-file')

const tail = new TailFile('/path/to/your/logfile.txt', {encoding: 'utf8'})
  .on('data', (chunk) => {
    console.log(`Recieved a utf8 character chunk: ${chunk}`)
  })
  .on('tail_error', (err) => {
    console.error('TailFile had an error!', err)
  })
  .on('error', (err) => {
    console.error('A TailFile stream error was likely encountered', err)
  })
  .start()
  .catch((err) => {
    console.error('Cannot start.  Does the file exist?', err)
  })

Example using pipe

This example is more realistic. It pipes the output to a transform stream which breaks the data up by newlines, emitting its own data event for every line.

const TailFile = require('@logdna/tail-file')
const split2 = require('split2') // A common and efficient line splitter

const tail = new TailFile('/path/to/your/logfile.txt')
tail
  .on('tail_error', (err) => {
    console.error('TailFile had an error!', err)
    throw err
  })
  .start()
  .catch((err) => {
    console.error('Cannot start.  Does the file exist?', err)
    throw err
  })

// Data won't start flowing until piping

tail
  .pipe(split2())
  .on('data', (line) => {
    console.log(line)
  })

Example using readline

This is an easy way to get a "line splitter" by using Node.js core modules. For tailing files with high throughput, an official Transform stream is recommended since it will edge out readline slightly in performance.

const readline = require('readline')
const TailFile = require('@logdna/tail-file')

async function startTail() {
  const tail = new TailFile('./somelog.txt')
    .on('tail_error', (err) => {
      console.error('TailFile had an error!', err)
    })

  try {
    await tail.start()
    const linesplitter = readline.createInterface({
      input: tail
    })

    linesplitter.on('line', (line) => {
      console.log(line)
    })
  } catch (err) {
    console.error('Cannot start.  Does the file exist?', err)
  }
}

startTail().catch((err) => {
  process.nextTick(() => {
    throw err
  })
})

Example for Clean Shutdown

TailFile will call flush() when quit() is called. Therefore, to exit cleanly, one must simply await the quit call. If the implementation wishes to keep track of the last position read from the file (for resuming in the same spot later, for example), then a simple listener can be added to always track the file position. That way, when quit() is called, it will get properly updated.

const TailFile = require('@logdna/tail-file')
let position // Can be used to resume the last position from a new instance

const tail = new TailFile('./somelog.txt')

process.on('SIGINT', () => {
  tail.quit()
    .then(() => {
      console.log(`The last read file position was: ${position}`)
    })
    .catch((err) => {
      process.nextTick(() => {
        console.error('Error during TailFile shutdown', err)
      })
    })
})

tail
  .on('flush', ({lastReadPosition}) => {
    position = lastReadPosition
  })
  .on('data', (chunk) => {
    console.log(chunk.toString())
  })
  .start()
  .catch((err) => {
    console.error('Cannot start.  Does the file exist?', err)
    throw err
  })

Events

TailFile is a Readable stream, so it can emit any events from that superclass. Additionally, it will emit the following custom events.

Event: 'flush'

• <Object>
  • lastReadPosition <Number> - The current file position at flush time

This event is emitted when the underlying stream is done being read. If backpressure is in effect, then _read() may be called multiple times until it's flushed, so this event signals the end of the process. It is used primarily in shutdown to make sure the data is exhausted, but users may listen for this event if the relative "read position" in the file is of interest. For example, the lastReadPosition may be persisted to memory or database for resuming tail-file on a separate execution without missing any lines or duplicating them.

Event: 'renamed'

• <Object>
  • message <String> - Static message: 'The file was renamed or rolled. Tailing resumed from the beginning.'
  • filename <String> - The filename that's being tailed
  • when <Date> - The date/time when the event happened

This event is emitted when a file with the same name is found, but has a different inode than the previous poll. Commonly, this happens during a log rotation.

Event: 'retry'

• <Object>
  • message <String> - Static message: 'File disappeared. Retrying.'
  • filename <String> - The filename that's being tailed
  • attempts <Number> - The number of attempts to try and access the file
  • when <Date> - The date/time when the event happened

If a file that was successfully being tailed goes away, TailFile will try for maxPollFailures to re-poll the file. For each of those retries, this event is emitted for informative purposes. Typically, this could happen if log rolling is occurring manually, or timed in a way where the poll happens during the time in which the "new" filename is not yet created.

Event: 'tail_error'

• <Error>
  • message <String> - The error message as written by TailFile
  • code <String> - Static value of ETAIL
  • meta <Object> - Additional metadata added for context
    • actual <Error> - The actual error that was thrown, e.g. from a stream

When an error happens that is specific to TailFile, it cannot emit an error event without causing the main stream to end (because it's a Readable implementation). Therefore, if an error happens in a place such as reading the underlying file resource, a tail_error event will be emitted instead.

Event: 'truncated'

• <Object>
  • message <String> - Static message: 'The file was truncated. Tailing resumed from the beginning.'
  • filename <String> - The filename that's being tailed
  • when <Date> - The date/time when the event happened

If a file is shortened or truncated without moving or renaming the file, TailFile will assume it to be a new file, and it will start consuming lines from the beginning of the file. This event is emitted for informational purposes about that behavior.

Event: (Any Readable event)

TailFile implements a Readable stream, so it may also emit these events. The most common ones are close (when TailFile exits), or data events from the stream.

API

Constructor: new TailFile(filename[, options])

• filename <String> - The filename to tail. Poll errors do not happen until start is called.
• options <Object> - Optional
  • pollFileIntervalMs <Number> - How often to poll filename for changes. Default: 1000ms
  • pollFailureRetryMs <Number> - After a polling error (ENOENT?), how long to wait before retrying. Default: 200ms
  • maxPollFailures <Number> - The number of times to retry a failed poll before exiting/erroring. Default: 10 times.
  • readStreamOpts <Object> - Options to pass to the fs.createReadStream function. This is used for reading bytes that have been added to filename between every poll.
  • startPos <Number> - An integer representing the inital read position in the file. Useful for reading from 0. Default: null (start tailing from EOF)
  • Any additional key-value options get passed to the Readable superclass constructor of TailFile
• Throws: <TypeError>|<RangeError> if parameter validation fails
• Returns: TailFile, which is a Readable stream

Instantiating TailFile will return a readable stream, but nothing will happen until start() is called. After that, follow node's standard procedure to get the stream into flowing mode. Typically, this means using pipe or attaching data listeners to the readable stream.

As the underlying filename is polled for changes, it will call fs.createReadStream to efficiently read the changed bytes since the last poll. To control the options of that stream, the key-values in readStreamOpts will be passed to the fs.createReadStream constructor. Similarly, options for controlling TailFile's' stream can be passed in via options, and they will get passed through to the Readable's super() constructor. Useful settings such as encoding: 'utf8' can be used this way.

tail.start()

• Returns: <Promise> - Resolves after the file is polled successfully
• Rejects: If filename is not found

Calling start() begins the polling of filename to watch for added/changed bytes. start() may be called before or after data is set up to be consumed with a data listener or a pipe. Standard node stream rules apply, which say that data will not flow through the stream until it's consumed.

tail.quit()

• Returns: <Promise> - Resolves after flush is called and streams are closed.
• Emits: close when the parent Readstream is ended.

This function calls flush, then closes all streams and exits cleanly. The parent TailFile stream will be properly ended by pushing null, therefore an end event may be emitted as well.

Program Flow

Using "file watcher" events don't always work across different operating systems, therefore the most effective way to "tail" a file is to continuously poll it for changes and read those changes when they're detected. Even Unix's tail -f command works similarly.

Once start() is called, TailFile will being this polling process. As changes are detected through a .size comparison, it uses fs.openReadStream to efficiently read to the end of the file using async/await iterators. This allows backpressure to be supported throughout the process.

How Log Rolling is Handled

TailFile keeps a FileHandle open for the filename, which is attached to an inode. If log rolling happens, TailFile uses the FileHandle to read the rest of the "old" file before starting the process from the beginning of the newly-created file. This ensures that no data is lost due to the rolling/renaming of filename. This functionality assumes that filename is re-created with the same name, otherwise an error is emitted if filename does not re-appear.

Backpressure Pauses Polling

Because TailFile won't be consumed until it is in a reading mode, this may cause backpressure to be enacted. In other words, if .start() is called, but pipe or data events are not immediately set up, TailFile may encounter backpressure if its push() calls exceed the high water mark. Backpressure can also happen if TailFile becomes unpiped. In these cases, TailFile will stop polling and wait until data is flowing before polling resumes.

Log Rolling During Backpressure

If polling is off during backpressure, TailFile can handle a single log roll or rename during backpressure, but if the log is renamed more than once, there will most likely be data loss, as polling for changes will be off.

This is an extrememly unlikely edge case, however we recommend consuming the TailFile stream almost immediately upon creation.

Contributing

This project is open-sourced, and accepts PRs from the public for bugs or feature enhancements. These are the guidelines for contributing:

• The project uses Commitlint and enforces Conventional Commit Standard. Please format your commits based on these guidelines.
• An issue must be opened in the repository for any bug, feature, or anything else that will have a PR
  • The commit message must reference the issue with an acceptable action tag in the commit footer, e.g. Fixes: #5

Changelog

4.0.2 (2024-04-11)

Bug Fixes

• ci: Skip release:dry for forks and call commitlint 9a26854 - Darin Spivey, closes: #46

Miscellaneous

• Fix docs on return value of quit 6491654 - Darin Spivey