Please use our updated library which supports specifying region to send your log data to.
Allows you to send logs to your logentries account from Node or io.js.
It might work with Browserify, too, but you would need to use shims for net and tls. Such shims do exist, based on forge, but I haven’t tested it. There’s a seperate client intended for use in the browser though, called le_js, which uses http and is optimized for browser-specific logging needs.
Tested in Node v0.10 + and io.js. It probably works in Node 0.8 too, but one of the test libraries (mitm) doesn’t, so it remains unconfirmed.
What is now "le_node" was previously "logentries-client"; users of le_node versions before 1.0.2 should read the sections below that detail differences if they wish to update.
var Logger = require('le_node');
var logger = new Logger({ token: 'myAccessToken' });
logger.warning('The kittens have become alarmingly adorable.')
The options object you provide to the constructor only requires your access token, but you can configure its behavior further.
All of the following except token, levels and secure can also be
configured after instantiation as settable properties on the client. They are
accessors, though, and invalid values will be ignored.
console.log,
console.warn and console.error as appropriate. Default: false.16192.false.100015 * 1000fibonacci or exponential. Default: fibonnacci15.15 * 60 * 1000.false. More details on this below.flatten is true, you can also indicate whether arrays
should be subject to the same process. Defaults to true if flatten is
true; otherwise meaningless.false.true.Error object, setting this to
true will cause the stack trace to be included. Default: false.secure is false, or 443 if
it’s true.true will enable debug logging with a default stdout
logger.log method.The default log levels are:
You can provision the constructor with custom names for these levels with either an array or an object hash:
[ 'boring', 'yawn', 'eh', 'hey' ]
{ boring: 0, yawn: 1, eh: 2, hey: 3 }
In the former case, the index corresponds to the numeric level, so sparse arrays are valid. In either case, missing levels will be filled in with the defaults.
The minLevel option respects either level number (e.g. 2) or the name (e.g.
'eh').
The level names each become methods on the client, which are just sugar for
calling client.log(lvl, logentry) with the first argument curried.
Since these names will appear on the client, they can’t collide with existing properties. Not that you’re particularly likely to try naming a log level ‘hasOwnProperty’ or ‘_write’ but I figured I should mention it.
So the following three are equivalent:
logger.notice('my msg');
logger.log('notice', 'my msg');
logger.log(2, 'my msg');
It’s also possible to forgo log levels altogether. Just call log with a single
argument and it will be interpretted as the log entry. When used this way, the
minLevel setting is ignored.
'error'The client is an EventEmitter, so you should (as always) make sure you have a
listener on 'error'. Error events can occur when there’s been a problem with
the connection or if a method was called with invalid parameters. Note that
errors that occur during instantiation, as opposed to operation, will throw.
'log'Triggered when a log is about to be written to the underlying connection. The prepared log object or string is supplied as an argument.
'connected' and 'disconnected' and 'timed out'These indicate when a new connection to the host is established, destroyed or timed out due to client side inactivity. Inactivity timeout is normal if the connection is inactive for a configurable period of time (see inactivityTimeout); it will be reopened when needed again. Disconnection can be either a result of socket inactivity or a network failure.
'drain', 'finish', 'pipe', and 'unpipe'These are events inherited from Writable.
'connection drain'DEPRECATED. Use buffer drain event instead.
'buffer drain'This event is emitted when the underlying ring buffer is fully consumed and Socket.write callback called. This can be useful when it’s time for the application to terminate but you want to be sure any pending logs have finished writing.
logger.notice({ type: 'server', event: 'shutdown' });
logger.once('buffer drain', () => {
logger.closeConnection();
logger.on('disconnected', () => {
process.exit();
});
});
'buffer shift'Buffer shift event is emitted when the internal buffer is shifted due to reaching bufferSize
of events in the buffer. This event may be listened for security/operations related reasons as
each time this event is emitted, a log event will be discarded and discarded log event will
never make it to Logentries.
logger.ringBuffer.on('buffer shift', () => {
// page devops or send an email
});
Log entries can be strings or objects. If the log argument is an array, it will be interpretted as multiple log events.
In the case of objects, the native JSON.stringify serialization is augmented in several ways. In addition to handling circular references, it will automatically take care of a variety of objects and primitives which otherwise wouldn’t serialize correctly, like Error, RegExp, Set, Map, Infinity, NaN, etc.
If you choose to set withStack to true, errors will include their stacktraces
as an array (so that they are not painful to look at). Be sure to turn on
"expand JSON" (meaning pretty print) in the options on logentries:
You can adjust this further by supplying your own custom replacer. This is a
standard argument to JSON.stringify -- See MDN: JSON > Stringify > The Replacer Parameter
for details. In the event that you supply a custom replacer, it is applied
prior to the built-in replacer described above so you can override its behavior.
Two options are available, timestamp and withLevel, which will add data to
your log events. For objects, these are added as properties (non-mutatively).
For strings, these values are prepended. If the name of a property would cause
a collision with an existing property, it will be prepended with an underscore.
In some cases it will end up being easier to query your data if objects aren’t
deeply nested. With the flatten and flattenArrays options, you can tell the
client to transform objects like so:
{ "a": 1, "b": { "c": 2 } } => { "a": 1, "b.c": 2 }If flattenArrays has not been set to false, this transformation will apply to
arrays as well:
{ "a": [ "b", { "c": 3 } ] } => { "a.0": "b", "a.1.c": 3 }In addition to log and its arbitrary sugary cousins, you can call
closeConnection to explicitly close an open connection if one exists; you
might wish to do this as part of a graceful exit. The connection will reopen if
you log further.
Also, because the client is actually a writable stream, you can call write
directly. This gives you lower-level access to writing entries. It is in object
mode, but this means it expects discreet units (one call = one entry), not
actual objects; you should pass in strings. This is useful if you want to pipe
stdout, for example.
If there’s a problem with the connection (network loss or congestion),
entries will be buffered in an internal ring buffer to a max of 16192(bufferSize)
entries by default. After that, internal ring buffer will shift records
to keep only last bufferSize number of records in memory. A log that indicates the
buffer was full will be sent to internal logger "once" this happens.
If console is true, these log entries will still display there, but they will
not make it to LogEntries.
You can adjust the maximum size of the buffer with the bufferSize option.
You’ll want to raise it if you’re dealing with very high volume (either a high
number of logs per second, or when log entries are unusually long on average).
Outside of these situations, exceeding the max buffer size is more likely an
indication of creating logs in a synchronous loop (which seems like a bad idea).
If the connection fails, it will keep retrying with a fibonacci backoff by default.
Connection retry will start with a delay of reconnectInitialDelay and the delay between each retry
will go up to a maximum of reconnectMaxDelay with each retry in fibonacci sequence.
Backoff strategy can be changed to exponential through constructor if necessary.
A connection to the host does not guarantee that your logs are transmitting successfully. If you have a bad token, there is no feedback from the server to indicate this. The only way to confirm that your token is working is to check the live tail on Logentries. I will investigate this further to see if there’s some other means with which a token can be tested for validity.
If Winston is included in your package.json dependencies, simply requiring the
Logentries client will place the transport constructor at winston.transports,
even if Winston itself hasn’t yet been required.
var Logger = require('le_node');
var winston = require('winston');
assert(winston.transports.Logentries);
When adding a new Logentries transport, the options argument passed to Winston’s
add method supports the usual options in addition to those which are Winston-
specific. If custom levels are not provided, Winston’s defaults will be used.
winston.add(winston.transports.Logentries, { token: myToken });
In the hard-to-imagine case where you’re using Winston without including it in
package.json, you can explicitly provision the transport by first requiring
Winston and then calling Logger.provisionWinston().
For Bunyan it’s like so:
var bunyan = require('bunyan');
var Logger = require('le_node');
var loggerDefinition = Logger.bunyanStream({ token: myToken });
// One stream
var logger1 = bunyan.createLogger(loggerDefinition);
// Multiple streams
var logger2 = bunyan.createLogger({
name: 'whatevs',
streams: [ loggerDefinition, otherLoggerDefinition ]
});
As with Winston, the options argument takes the normal constructor options (with
the exception of timestamp, which is an option you should set on Bunyan itself
instead). Bunyan uses six log levels, so the seventh and eighth, if provided,
will be ignored; by default Bunyan’s level names will be used.
The object returned by bunyanStream is the Bunyan logging ‘channel’ definition
in total. If you want to futz with this you can -- you can change its name or
get the stream object itself from here.
When you create an account at Logentries (just a standard signup form; there’s a free tier), you can find the token you need. It’s shown during the initial walk- through but you can find it later under Logs/Hosts/{ the name of your host } -- on the far right, a gray TOKEN button that you can click to reveal the string.
That’s it -- once you have the token you’re set.
Previously, "le_node" and "logentries-client" were two different modules. The former has been replaced with the latter codebase, but the le_node name is the canonical repo (it’s referenced in many places). It’s still possible to get logentries-client under that name on NPM, but it’s soon just going to be an alias for this repository, le_node.
For users of le_node from before this switch, there are some important differences to note before upgrading.
The new codebase does follow the same essential pattern. If you only used the client constructor and the log methods previously, there may be no breaking changes for you. But for anybody else...
client.end()Unlike old le_node, the client is itself a writable stream (and therefore you
can pipe to it, for example from stdout, though note that 1 write invocation =
1 log entry). This also means that it has standard writable stream events and
methods, including .end(). In the old le_node, .end() was a non-stream
method that closed the underlying connection to the host.
For the functionality previously provided by .end(), use .closeConnection().
client.level() and client.winston()The old le_node had a method called level() for setting the minimum log level.
This is now a property (not a method) called minLevel. It can be set to either
the name of the level or its index. The level() method has been added to the
new codebase to facilitate migration, but will be removed at a later date.
Simply requiring le_node now automatically provisions Winston, if present, with
a Logentries transport constructor. You don’t have to do anything else. The
winston() method has also been added to the new codebase to prevent errors,
but it’s a noop and will be removed at a later date.
The old documentation seemed to suggest that placing a listener on the client for error events was an optional thing. This isn’t the case (and wasn’t the case in the old client, either). An unhandled error event from an EventEmitter is an unhandled error period. If you don’t place a listener for error events, your application will crash if the client emits an error.
The new codebase has a lot of new features, including some that are similar to, but not necessarilly the same as, old features that had been removed at some point or were just not documented.
The outstanding issues that exist for le_node at the time of writing are mostly things which either never affected this codebase or no longer make sense in regard to it.
time and level properties will never collide with existing props and are
both optionalYou should assume that there are other breaking changes which I am unaware of. When I wrote Logentries Client I hadn’t considered that it might replace le_node, so unfortunately interoperability was not on my mind. You’ll wish to test thoroughly before updating an existing codebase to use the new client.
.end() method will not synchronously terminate the underlying
connection anymore. Instead, it waits for drain before doing so.level property.level() and winston() methods with deprecation warnings so that
existing le_node applications do not throw TypeErrors.withLevel and timeout optionshost and port options for testingclient.logclient.log method no longer accepts an arbitrary number of
log entry arguments (to support above case, which seems much likely to be
useful)minLevel value to zero (1 was an accident)flatten and flattenArray optionswithStack set to truewithStack optionrunInContext() to prevent template string problems(Pieced together to the best of my ability by reviewing commit history.)
Please email our support team at support@logentries.com if you need any help.
FAQs
Logentries client for node; use with or without Winston or Bunyan.
We found that le_node demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 4 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Research
A malicious npm package disguised as a WhatsApp client is exploiting authentication flows with a remote kill switch to exfiltrate data and destroy files.
Security News
PyPI now supports digital attestations, enhancing security and trust by allowing package maintainers to verify the authenticity of Python packages.
Security News
GitHub removed 27 malicious pull requests attempting to inject harmful code across multiple open source repositories, in another round of low-effort attacks.