Posts by Nicholas Blumhardt

Filtering events on arrival at Seq

Collecting too much log data makes the “needle in the haystack” all the more difficult to find, but for practical reasons it is often easier to collect as much data as possible anyway and “sort it out later” via Seq retention policies.

These have served us well since the early versions of Seq, and selective retention is a feature that’s been widely appreciated by users. Retention policies do come at a cost however – as events are removed from the stream, space in the storage engine needs to be compacted and this can use significant resources on the server (CPU, RAM and I/O).

In many cases, it is preferable to drop uninteresting events as they arrive. This facility is now available in Seq 1.4.2-pre as “input filters”.

Each Seq API key can now specify a filter expression – just like the event stream supports – that identifies events that will be accepted when logged with the key:


A typical use case may be to drop events below a certain level, but any criterion can be included.

If for any reason it becomes necessary to change the filter during operation, updating it takes immediate effect so that more or fewer events can be recorded.

It’s a simple feature right now that we’re looking for feedback on – if you can grab the new preview from the download page (preview link is at the bottom) and try it out it would be great to hear your thoughts!

Not currently in the preview, but also coming in 1.4 is a MinLevel() query function that will tidy up filtering by level like the example shows.

Durable log shipping, from Serilog to Seq

A few days ago I took an impromptu snapshot of one of the Azure machines I have running Seq. Cloud storage being what it is, an I/O problem during the snapshot process crashed the Seq service, but I didn’t notice until the following day. The machine was running horrifically slowly at this point, so I restarted it and all came good.

When I logged into Seq to check some queries, I expected to see a big gap during the outage. It took me a few seconds to realize why there wasn’t one – all of the events raised during the outage period were there! – and that made me think this feature of the Seq client for Serilog might not be as well-known as I’d assumed.

When you set up Serilog to write events to Seq, you use code typically like:

Log.Logger = new LoggerConfiguration()

This will send events in batches directly using HTTP; network problems and hard app crashes can result in event loss.

Some time ago, another parameter became available:

Log.Logger = new LoggerConfiguration()
  .WriteTo.Seq("http://my-seq", bufferBaseFilename: @"C:\Logs\myapp")

If a buffer location is specified, the Seq client will write events to a set of rolling files like "myapp-2014-06-27.jsnl" (newline-separated JSON) and then ship them to the server when it’s possible. Almost nothing to configure, and voila! Reliable log collection.

There’s still no substitute for a redundant, transactional data store for truly critical events: machines can be lost, disks fill up and so-on. For general application monitoring and diagnostics however, I’m really pleased with how much “bang for the buck” this feature provides.

Can we change the face of JavaScript logging, too?

If you use Serilog in .NET and write code for the web, chances are you’ve wondered whether similar hybrid “structured/text” loggers exist for the browser and Node.js


If you haven’t tried Serilog, here’s a one-minute primer (using JavaScript syntax) so you can follow the rest of this post. Using Serilog is similar to every logging framework, ever:

var name = "Nick";
var wait = 45;
log("Hi {name}, your flight leaves in {wait} minutes", name, wait);

Look carefully at the string formatting though – particularly the way we give names to each of the parameters. The “message” part isn’t destructively rendered into text, but preserved as an object with property values. It can be formatted to the console etc. as usual:

[inf] Hi Nick, your flight leaves in 45 minutes

But more interestingly, it can be serialized in a format like JSON for processing and querying:

  "message": "Hi Nick, your flight leaves in 45 minutes",
  "messageTemplate": "Hi {name}, your flight leaves in {wait} minutes",
  "properties":  {
    "name": "Nick",
    "wait": 45

Preserving the structure of log data, and making it trivially easy to enrich that data with more contextual information, makes it possible to do some really clever things with the resulting logs.

Try finding all messages where the user’s name starts with “N”, or where the wait time is greater than one hour in a traditional log file if you’re skeptical here. In some trivial cases a regular expression might handle this cleanly, but it quickly becomes complex and cumbersome when more data is involved. Using Serilog it can be as simple as startswith(name, "N") or wait > 60.

Preserving the message template itself ("Hi, {name}...") also has some interesting and very useful results: the message template becomes a kind of “message type” that can be used to find – or exclude – messages of a specific type. Again very hard, often impossible with traditional loggers.

What’s going on in JavaScript?

In JavaScript as elsewhere, there are already some sophisticated logging frameworks, with varied levels of structured data support. A fairly new project called node-bunyan interests me the most, with its clever DTRACE integration and support for serializing objects instead of messages, but even here the new hybrid-style API and DSL provided by Serilog is missing. With node-bunyan it’s possible to log an object and/or a message, but messages are still formatted out in a lossy manner like the earlier loggers on .NET did.

Whether it grows into a standalone project in its own right, or encourages more established frameworks to take a look at this new style of logging, bringing Serilog’s API to JavaScript seems like a fun and worthwhile thing to try.


Tinkering away over a month of so I’ve sketched out what I think serilog.js should look like.


Yes, serilog.js terminal output is type-aware, just like syntax highlighting in an IDE! It’s useful when dealing with structured events to see the difference between a number 5 (magenta for numbers) and a string "5" (cyan for strings). In itself this feature is a nice trimming, but it’s not just a party-trick; below the surface, the event is carrying all of this information.

Getting started – installing from NPM

All the code is currently published to NPM. At the command line:

npm install serilog

Then in your node application:

var serilog = require('serilog');
var terminal = require('serilog/src/serilog-terminal-sink');
var file = require('serilog/src/serilog-file-sink');

Serilog “sinks” are destinations for log events. This example uses the terminal sink, which writes to the system terminal/console. At the moment, the sinks are all zipped up directly in the serilog package, but they need to be split out.

Tour of the API

At present, with very little of the back end ‘sinks’ built, serilog.js is all about the API.

Configuring the pipeline

One of the first things to notice is that the logger is a pipeline. Thinking of log events as structured data leads naturally to an API more like what you’d find in event processing frameworks than the typical logging library.

var log = serilog.configuration()
  .writeTo(file({path: 'log.txt'}))

Each element is applied in turn:

  • serilog.configuration() kicks things off by creating a new configuration object
  • .writeTo(terminal()) emits all events to the terminal sink
  • .minimumLevel('WARNING') filters out any that are below the WARNING level (we’ll cover levels below)
  • .writeTo(file({path: 'log.txt'}) emits the events that reach it to a log file, and
  • .createLogger() takes all of these pipeline stages and bundles them up into an efficient logger object

The result is log, which we’ll use below. (As in Serilog for .NET, there’s no “global” logger – multiple loggers can be created and are completely independent, unless you link them together with something like .writeTo(log2)).

You might be wondering – why not Node.js streams or RxJS etc.? Simple really – generic APIs are great for glueing together solutions, but for something that’s used extensively in a codebase, maintaining complete control over dependencies, performance and semantics is a good idea long-term, and that’s the path taken here.

Logging levels

serilog.js has four simple logging levels, listed here in order of importance:

log.trace('This is a lot of detail');
log.information('Something of note happened');
log.warning('Look out, wheels are coming off!');
log.error('Something's broken');

If log.information() looks like a bit of a mouthful, don’t worry: the log object itself is a function that can be called as a synonym for information:

log('This is an information-level event');

The default level for a new logger is INFORMATION; using .minimumLevel('TRACE') at the start of the pipeline will bump this up.

Message templates

The syntax of message templates is carried over directly from Serilog. Event properties are named by including them in braces, and the values are provided as positional arguments:

log('{a} + {b} = {c}', 1, 2, 3);
// -> {properties: {a: 1, b: 2, c: 3}}

I’ve left the message/template/timestamp out of the example output (// -> ...) to keep the focus on the structured part of the event, but these other parts of the event exist too.

If you pass an object to the log, e.g.:

var mail = {to: 'nblumhardt', subject: 'Is the post finished?'};
log('Sending {mail}', mail);
// -> {properties: {mail: '[Object] object'}}

You might be surprised that it’s stored using toString(). We haven’t missed a great opportunity to serialize here – the problem is that it is very easy to serialize too much data by mistake this way.

If you know that what you’re passing into a log method is a nice little chunk of data to work with later, prefix the property name with @ to say “I know what I’m doing – serialize this!”:

var mail = {to: 'nblumhardt', subject: 'Is the post finished?'};
log('Sending {@mail}', mail);
// -> {properties: {mail: {to: 'nblumhardt', subject: 'Is the post finished?'}}}

serilog.js can handle structured storage of objects and arrays nicely, and these can support some really cool processing scenarios – just don’t forget the @ (if you watch the events roll by on the terminal, you’ll notice the difference too – objects are colored green, while strings render in cyan).

Context and enrichment

The strength of structured logs is in correlation; this is one of the great challenges when instrumenting distributed apps, and probably the #1 reason that interest in structured logs has picked up again in recent times.

When searching a large number of events produced across many machines, it’s vital to have properties to filter on so that a single location or thread of interaction can be followed.

In a simple case, you might want to instrument all events with the name of the machine producing them. This can be done when configuring the logger:

var log = serilog.configuration()
  .enrich({machine: env.HOSTNAME})

Events created by the logger will all carry the specified properties:

log('Hello, {name}', 'Nick');
// -> {properties: {name: 'Nick', machine: 'NBLUMHARDT-RMBP'}}

More often though, it’s desirable to tag events produced within a part of the application, or in a transaction. For this the using() method is provided on the logger itself:

var smtpLog = log.using({subsystem: 'SMTP'});
smtpLog('About to send mail to {name}', 'Nick');
// -> { properties: {subsystem: 'SMTP', name: 'Nick'}}

A final part of the API to check out is filtering. While most filtering will happen server-side wherever you collect your logs, it can be useful to apply some filters in the logging pipeline if you have a component producing a lot of noise, or if you need to avoid passing events with sensitive data to logs.

This fits in as you’d expect:

var log = serilog.configuration()
    return evt.messageTemplate.raw !== 'Making {noise}' && !== 'SMTP';
Building serilog.js

If you want to tinker with the project, you can grab the source from GitHub. The build process uses gulp to drive some mocha tests. Just type:

npm install

…and you’re away!

The browser

serilog.js also runs in the browser (though version support will be patchy and is completely untested!). There are some notes on the GitHub site linked above on how to do this, and a test.html page in the repository that sets some of it up.

The catch

Serilog in .NET is out there in production use today. It’s being extended and refined by a core group of contributors and takes what time I have available to coordinate.

I’ll be using serilog.js in a few small JavaScript projects and extending it for my own needs, and I’ll be building a sink that writes to Seq at some point. But, I don’t have the capacity to really drive this project beyond the prototype that it is right now.

If you’re interested in JavaScript, event processing, application monitoring and/or API design, serilog.js needs you. Kick the tyres, check out the code, send a pull request! It’s a project “up for grabs” in the largest sense. Can you help change the face of JavaScript logging?

Seq 1.4 preview is out!

The features planned for Seq 1.4 are starting to come together. There’s a new preview available to download on the Seq website (direct download).

While 1.4 is still a little way off “done”, many of the improvements we’ve made are usable today; here’s a run through what you can expect from the new version.

What’s Seq? Seq is the fastest way for .NET teams to improve visibility using structured application logs. If you’re using Serilog, you’ll find Seq is the perfect partner – trivially easy to set up, and built for structured log events so querying them is simple from “day 1” of a project. You already use CI and automated deployment: Seq is the next piece of the puzzle, closing the loop to show your team what’s happening in development, testing and production.


New in 1.4:


Charts show the count of matching events in a time period. You can view the last day (by hour), last week (by day), last month (by day) or last year (by week). It’s a simple feature that I’m sure we’ll expand on, but already a great way to keep a “finger on the pulse”.

To add a chart, filter the “events” view to show just what you want to chart, then drop down the “refresh” button and select “Add to dash”. By default, queries are shown on the dash as counters; to toggle between the counter and chart display, there’s a little icon in the top left that appears on hover.


(If you’re already using counters with Seq 1.3, you’ll need to recreate them before historical data will be indexed by the 1.4 preview.)

Quick(er) switching between views

A much-requested improvement: when no view is selected, the first 10 views will be shown in a list (saving a few clicks when switching between views).


Extended text operators

A bunch of work has gone into better string comparison functions in this version. Seq queries now support:

  • Contains(text, pattern)
  • StartsWith(text, pattern) and EndsWith(text, pattern)
  • IndexOf(text, pattern), and
  • Length(text)

In each of these, “pattern” can be case-insensitive, case-sensitive, or a regular expression. There’s some more information and examples in the documentation.

Date functions and @Timestamp

Events in Seq now carry a queryable @Timestamp property, and we’ve added a DateTime() function that can be used with it. Check out the documentation for some examples here.

Keep the feedback coming!

Download the preview, try it out, and let us know what you think! You know where to find me :) – and there’s also a Seq discussion group we’d love you to join, and a Seq Twitter account if you’d like to keep in the loop.

Serilog gets a mention at NDC Oslo 2014

It was really cool to see this session from NDC about Elasticsearch/Logstash/Kibana featuring Serilog from just after 16:00.

Monitoring your app with Logstash and Elasticsearch from NDC Conferences on Vimeo.

It’s fun watching Serilog presented by people who obviously “get it”. And, compared with the alternatives shown earlier in the session, it’s great to see just how easy Serilog is for generating quality structured logs.

(If you’re not using Serilog and interested in the demo code, it’s on GitHub, but please check out this tiny PR if it hasn’t made it through yet.)

Serilog output template formatting changes

The latest Serilog version 1.3.7 fixes some long-standing text formatting quirks.

Serilog has a bit of an idiosyncratic take on formatting text. You could probably say the same about its designer: before Serilog, I used to write a lot of log statements like this:

Log.Information("Could not find documents matching '{0}'", searchTerm);

If you look carefully you’ll note an extra couple of single quotes (') in there around the the {0} token that will be replaced with the search term being logged. I got into this habit because when reading logs I find it less jarring to see random string data quoted, so that my brain doesn’t try to join the whole lot up in a single bewildering sentence:

Could not find any documents matching my pants


Could not find any documents matching 'my pants'

String formatting in Serilog message templates

When I started writing Serilog I thought I’d save a few characters, and encourage consistency, by making this little idiom part of its string formatting engine:

Log.Information("Could not find documents matching {Term}", searchTerm);


Could not find any documents matching "my pants"

I’m sure there are plenty of people who’d rather a different default, but so far no one’s raised it as particularly problematic. If you need to match a specific output format you can use a special “formatting” specifier to turn this off – just append :l (yes, a lowercase letter L) to the token:

Log.Information("Could not find documents matching {Term:l}", searchTerm);


Could not find any documents matching my pants

Reusing message templates for output

The formatting above is applied when rendering log messages into text. A bit later in the design process, I think some time around when Serilog became a public project, we needed another formatting mechanism, this time to control how the different parts of the log event including the message get laid out e.g. for display at the console.

Log.Logger = new LoggerConfiguration()
  .WriteTo.Console("{Timestamp:G} [{Level}] {Message}{NewLine:l}{Exception:l}")

You’ll notice the same style of format string appears in there. The engine used to write the text is the same; there were (and are) a lot of advantages to reusing this code for “output templates” like this.

Unfortunately, some of the decisions that work well when rendering a structured message, get in the way when presenting text like this. You’ll see above one of the oddities – the pre-defined NewLine and Exception properties are represented as strings under the hood, so the :l literal specifier has to appear in there to avoid unsightly quotation marks appearing in the output. Not a big deal really, but too much to have to grasp as a new user.

The other issue is that in message templates, i.e. when calling Log.Information() and friends, it’s an error to provide less arguments than the number of tokens that appear in the string. Serilog helpfully prints out the whole token as text – {Token} so that these issues are obvious when reading the rendered message.

In output templates this isn’t so desirable. Using {HttpRequestId} in the format string should just be ignored if the event doesn’t have such a property attached.

Serilog 1.3.7 changes

I almost versioned this one “1.4”, given there’s quite a change being made, but since the API is binary-identical this one’s out as a patch.

Put simply, Serilog now has slightly different handling of message templates vs. output templates. Message templates keep their current behaviour – quotation marks and all. When used to configure a sink as an output template however:

  • The :l specifier is now “on by default”
  • Missing tokens will simply be ignored

The default output templates all used :l formatting anyway, and so will almost all custom ones, so for 99% of current users I’d say this change will go unnoticed. If you define your own output templates though you can now clean up a bit of clutter and remove the literal specifier from fields like {Exception} and {NewLine}.

If you’re coming to Serilog as a new user, hopefully these changes give you one less reason to go looking for explanatory posts like this one!

Logging and locked files

Serilog 1.3.5 includes some important improvements to the rolling file sink.

No doubt the least pleasant aspect of dealing with log files is concurrency. When multiple processes write to the same log file, either:

  1. They need to coordinate writes using system-wide mutual exclusion (unreliable)
  2. They need to wait for the file to become “free” and close it immediately after writing (slow)
  3. One of them must fail (inconvenient)

Serilog takes a bit of a stand on this and chooses option 3), requiring that each process uses its own log file, or set of rolling files.

This has the advantages of predictability and performance, and I’m happy with the choice.

But, there’s a catch, and it is a nasty one.

IIS and overlapped application pool recycling

When an IIS app pool is “recycled” (shut down and re-started to clean up any leaky resources) the IIS pipeline’s pretty smart about it. Rather than leave a site offline between unloading the old worker process and loading the new one, IIS starts the new process first and only unloads the old one once the new one is serving requests.

Unfortunately, this means the new app pool needs to open a log file still locked by the outgoing process. No logs for you!

Disabling overlapped recycling

For typical intranet sites, or web-facing sites behind a load balancer, disabling overlapped recycling isn’t a huge issue. To date this has been the solution for using Serilog with IIS. But, it isn’t an option for some, and happily those “some” were loud and determined enough to get a better solution into Serilog 1.3.5.

The old rolling scheme

Serilog still doesn’t handle locked single-file logs. Generally these kinds of logs aren’t used in e.g. IIS sites, so the old caveats of a file-per-process still apply if you use WriteTo.File().

Rolling files however already provide a model that maps quite naturally to the “rolling over” of IIS worker processes.

A sequence of files written by the rolling file sink might look like:


As the days pass a new file is created, and eventually the old one is removed.

The fix: rolling when a file is locked

Now, assuming that it’s the third of May, if the IIS app pool is recycled you’ll see a sequence like:


The _001 file will be used by the new worker process, while the old one hangs on to the non-prefixed file.

Some guarantees

What happens when the new process, using the suffixed file, is recycled?

Well, the simplistic response would be for the replacement to open the non-suffixed file, since that is now in all likelihood unlocked. But this would make reading the files a pretty awful experience, as date/time ranges jumped around between them.

Instead, the new process will create:


Important notes

The only caveat to note here is that the implementation is robust to the scenarios that commonly occur when restarting/recycling processes. It doesn’t address long-lived concurrent processes, for that you’ll need to decide on an alternative scheme (or use a more concurrency-friendly sink than what the file sinks can provide).

You should also be aware that the retained file count provided by the rolling file sink is per-file, not per-day, so if you expect frequent overlapped log file recycling, you should allow a few more retained files to ensure you keep the required date range. We’ll probably implement a retained-days count in the future.

Is this the long-term plan?

This is the long-term solution to locked files in Serilog, at least as far as any current plans cover. How does it work for you? Serilog’s a community project and always open to constructive feedback and good ideas.


Seq 1.3 – what’s new?

Seq 1.3 is now available at Since version 1.2 we’ve introduced the ‘dash’ – a hub for your most-used queries and bookmarked events – and made refinements to many other parts of the app.


Seq is designed for keeping tabs on your running applications. For the really important events, you’ll probably use email alerts to be notified when something needs attention, but if you’re like me you’ll have your finger on the pulse of quite a few things, not all of them critical.

For that, Seq 1.3 introduces watches – a saved view over the event stream that Seq will monitor for new events. When an event matches a watch, after a short delay you’ll see that on the shiny new dash:

Dash 1.3

Clicking through to view the matching events will reset the watch counter to zero.

To start watching a view, query, filter or combination of those things, drop down the Refresh button and select Add watch.



Alongside watches on the dash you’ll also find bookmarks. A bookmark is just a link to an individual event, but for as long as the event is bookmarked it won’t be cleaned up by any retention policies you might have set.

Bookmarks are added from the new Pin menu that appears on the detail view of each event.



You can also bookmark an event for another member of your team by choosing the Assign to another user item under Pin.


This is called an assignment and will show up, with some notes if you include them, on the other person’s dash.


If you need to link an event to send it by email etc. you can still create a permalink – these are what they imply, and now also save the event from any applicable retention policy.

Causal ordering

Hiding away in Seq 1.3 is a new killer feature: the “arrival ordering” queries in Seq 1.3 make debugging causality easier and more fun.

Let’s say we’ve hit on an interesting event: a crash or something out of the ordinary that we need to investigate.


In the event details, the new Id drop-down includes Find with predecessors. This will restrict the event stream so that the event we’re viewing is at the top.

Now, by requiring this filter in the query (press Require) we can slice the predecessors of the event along whatever axes we like. What was happening on the machine before this event?


Whatever filter we add, the results will be shown up to and including the event we chose. We can find the last error logged anywhere before this one; the last event logged for the same customer; the last events in the same web request…

Tedious questions to try to answer with a flat log file, but now an absolute breeze with Seq.

Offline compaction

A side effect of the current storage engine Seq uses is that sometimes fragmentation may cause the event store to take more space on disk than it needs. We’re digging deeply into the storage subsystem in coming releases and will ideally provide full online compaction, but in the interim if you find your event store is disproportionately large, you can run:

seq.exe stop
seq.exe compact
seq.exe start

…to reclaim some space.

Virtual directory support

Seq has always allowed listen URLs with deep paths, however in past versions some bugs in the UI meant these didn’t fully work.

In Seq 1.3 you can now set up multiple instances on the same server at nested paths, e.g.


To find out more about setting up instances see the documentation.

Help has moved

We’re putting all of our documentation efforts into improving so it feels only right that help in the application will take you there.

But there’s more!

  • Boolean queries and comparisons with null now match their documented behaviour
  • The installer handles reconfiguring service paths
  • The filter ‘clear’ button is now clickable even with focus on the input field
  • IE ‘compatibility mode’ is now avoided when Seq is run as an intranet site
  • @Id can now be used in filters
  • The ‘single event view’ now correctly supports further filtering on properties of the displayed event
  • Filter history is auto-completed within a session
  • JSON formatting is improved

Actually, it’s hard to capture just how much has changed in the last short month of development. Give Seq 1.3 a try – we think you’ll enjoy the improvements!

Download Seq 1.3 from the project site. The very handy Developer edition is free, and you can request a 30-day trial of Seq Enterprise if you would like to try out multi-user support.

One year of Serilog

I just pushed the first build of Serilog 1.3 to NuGet. This one includes a more flexible JsonFormatter and support for width specifiers in format strings. Not huge changes, but both called for obsoleting some existing methods, and with 1.2 at the venerable age of “.53″ it was time to rev and get back to the nice simple version “1.3.1”.

It occurred to me that we’ve now hit the one year mark in the Serilog project. Version 0.1 was announced on March 29th, 2013 actually!

If there was any doubt back then of the viability of “yet another logger” in the .NET ecosystem, it’s been well and truly dispelled in my mind. Sixteen of us have now contributed to the project, and what’s most exciting to me is that the developers extending, using and discussing Serilog are some of the best and brightest I know. It’s the kind of momentum that makes me confident we’ll see more use and growth in the next year.

For those who follow it, the mention of structured logging as a technology to “trial” in the January 2014 ThoughtWorks Technology Radar is also a good indicator that Serilog is aligned with the facilities .NET developers will come to expect and use widely.

Our 1.0 came only six months ago, so these are still early days, but there are now 38 packages tagged “serilog” on NuGet, and despite the more-than-occasional glitches in the counter ;) the core Serilog package has over 7,400 downloads.

Visits to are also looking healthy:


April’s been a visibly good month, with a lot of interest stirred up by a discussion on .NET Rocks! where I did my best to “get the word out” about how Serilog has totally transformed my thinking about logging.

If you’re using Serilog it would be great to hear more about how it has worked for you, and where you think the project should be aiming next – get in touch here or on our discussion list!

XML configuration for Serilog

Some issues hang on! Serilog’s #3 — XML configuration support — was still grimacing at us from the backlog alongside issues with three digits, until it was unceremoniously closed a few minutes ago.

If you’ve spent any time with Serilog you’ll have noticed the absence of built-in XML configuration support. Something about the project just wanted – and wants – to be a code-centric thing. I guess it is part of lifting the status of logging a bit – ORMs for example moved on from XML configuration to code-first/fluent configuration many moons ago. Instrumenting your app with an event stream is an important thing that deserves a modern configuration API.

XML’s not at all bad though. File paths change, servers have different addresses in different environments, logging levels go up and down. XML’s a great fit for this and we’ve known that all along with Serilog; we’ve just chosen to keep it in its place and write code like:

var logFile = ConfigurationManager.AppSettings["LogFilePath"];

var logger = new LoggerConfiguration()

Over time this gets boring and repetitive I admit. Thats why we’ve now got the Serilog.Extras.AppSettings package.

XML configuration with a twist

As I said – we like XML configuration in its place, but that’s a pretty small place. When I open up App.config or Web.config in a modern .NET project the amount of gunk in there can be pretty confronting.

Most XML config implementations provide their own <configurationSection>s, and those need to be declared, and then the XML syntax used within them generally gets copied from a long list of arcana like the one in the log4net documentation that I’ve visited for years on end.

Serilog’s “app settings” support is different; we use a technique I first saw suggested for Autofac module configuration several years ago.

This revolves around the <appSettings> collection, an element that’s present in almost every .NET configuration file. Here’s an example to illustrate what we’re doing:

  <add key="serilog:minimum-level" value="Verbose" />

Instead of our own custom <serilog> element, we cheat a bit and just look for <appSettings> elements that match a certain pattern (their names start with serilog:). Simple idea, but effective: I’m going to write this whole article without checking the documentation or sample code even once! Remembering a setting name or pattern is much easier for me than remembering a whole pile of custom XML syntax.

Going a bit further, here’s the configuation for two sinks, the console and a log file:

  <add key="serilog:write-to:ColoredConsole" />
  <add key="serilog:write-to:File.path" value="C:\Logs\myapp.txt" />

(Notice you don’t actually have to provide a value for an app setting? I didn’t until recently, but works nicely!)

Where do the magic names ColoredConsole and File.path come from? They’re just encodings of the following method calls:

  .WriteTo.File(path: @"C:\Logs\myapp.txt")

The name of the path parameter obviously wouldn’t be specified in normal usage, but you can see how the XML and code-based configuration match up 1:1.

The extension methods are looked up dynamically using reflection, so you can configure your own sinks this way. You can include multiple parameters for the same sink using multiple keys – for both features check out the wiki page linked below.

Enabling the package

Two steps: 1. Install the Serilog.Extas.AppSettings package from NuGet; 2. add a call to ReadAppSettings() when you’re configuring your logger:

var logger = new LoggerConfiguration()
  ... // Other configuration here, then

You can mix and match app settings with code-driven config, but for each sink you need to use one or the other.

There’s much more detailed documentation here, on the wiki.

But isn’t this super-limited?

I’m happy to trade configuation power for something I can actually remember how to use. Someday we might have a more sophisiticated XML configuration option than this, but unless you can write the PR – I’m served pretty well by this one and not in a hurry.

But this is totally not my style!

That’s fine; it’s an ‘extra’, meaning we ship it separately from Serilog’s core and you’re free to install it if you like it and ignore it otherwise.

Still, I’ll be happy to hear if you do find it useful!