Scott Hanselman

A complete containerized .NET Core Application microservice that is as small as possible

September 14, 2018 Comment on this post [4] Posted in Docker | Open Source
Sponsored By

OK, maybe not technically a microservice, but that's a hot buzzword these days, right? A few weeks ago I blogged about Improvements on ASP.NET Core deployments on Zeit's now.sh and making small container images. By the end I was able to cut my container size in half.

The trimming I was using is experimental and very aggressive. If you app loads things at runtime - like ASP.NET Razor Pages sometimes does - you may end up getting weird errors at runtime when a Type is missing. Some types may have been trimmed away!

For example:

fail: Microsoft.AspNetCore.Server.Kestrel[13]
Connection id "0HLGQ1DIEF1KV", Request id "0HLGQ1DIEF1KV:00000001": An unhandled exception was thrown by the application.
System.TypeLoadException: Could not load type 'Microsoft.AspNetCore.Diagnostics.IExceptionHandlerPathFeature' from assembly 'Microsoft.Extensions.Primitives, Version=2.1.1.0, Culture=neutral, PublicKeyToken=adb9793829ddae60'.
at Microsoft.AspNetCore.Diagnostics.ExceptionHandlerMiddleware.Invoke(HttpContext context)
at System.Runtime.CompilerServices.AsyncMethodBuilderCore.Start[TStateMachine](TStateMachine& stateMachine)
at Microsoft.AspNetCore.Diagnostics.ExceptionHandlerMiddleware.Invoke(HttpContext context)
at Microsoft.AspNetCore.HostFiltering.HostFilteringMiddleware.Invoke(HttpContext context)
at Microsoft.AspNetCore.Hosting.Internal.HostingApplication.ProcessRequestAsync(Context context)
at Microsoft.AspNetCore.Server.Kestrel.Core.Internal.Http.HttpProtocol.ProcessRequests[TContext](IHttpApplication`1 application)

Yikes!

I'm doing a self-Contained deployment and then trim the result! Richard Lander has a great dockerfile example. Note how he's doing the package addition with the dotnet CLI with "dotnet add package" and subsequent trim within the Dockerfile (as opposed to you adding it to your local development copy's csproj).

I'm adding the Tree Trimming Linker in the Dockerfile, so the trimming happens when the container image is built. I'm using the dotnet command to "dotnet add package ILLink.Tasks. This means I don't need to reference the linker package at development time - it's all at container build time.

FROM microsoft/dotnet:2.1-sdk-alpine AS build
WORKDIR /app

# copy csproj and restore as distinct layers
COPY *.sln .
COPY nuget.config .
COPY superzeit/*.csproj ./superzeit/
RUN dotnet restore

# copy everything else and build app
COPY . .
WORKDIR /app/superzeit
RUN dotnet build

FROM build AS publish
WORKDIR /app/superzeit
# add IL Linker package
RUN dotnet add package ILLink.Tasks -v 0.1.5-preview-1841731 -s https://dotnet.myget.org/F/dotnet-core/api/v3/index.json
RUN dotnet publish -c Release -o out -r linux-musl-x64 /p:ShowLinkerSizeComparison=true

FROM microsoft/dotnet:2.1-runtime-deps-alpine AS runtime
ENV DOTNET_USE_POLLING_FILE_WATCHER=true
WORKDIR /app
COPY --from=publish /app/superzeit/out ./
ENTRYPOINT ["./superzeit"]

I did end up hitting this bug in the Linker (it's not Released) but there's an easy workaround. I just need to set the property CrossGenDuringPublish to false in the project file.

If you look at the Advanced Instructions for the Linker you can see that you can "root" types or assemblies. Root means "don't mess with these or stuff that hangs off them." So I just need to exercise my app at runtime and make sure that all the types that my app needs are available, but no unnecessary ones.

I added the Assemblies I wanted to keep (not remove) while trimming/linking to my project file:

<Project Sdk="Microsoft.NET.Sdk.Web">

<PropertyGroup>
<TargetFramework>netcoreapp2.1</TargetFramework>
<CrossGenDuringPublish>false</CrossGenDuringPublish>
</PropertyGroup>

<ItemGroup>
<LinkerRootAssemblies Include="Microsoft.AspNetCore.Mvc.Razor.Extensions;Microsoft.Extensions.FileProviders.Composite;Microsoft.Extensions.Primitives;Microsoft.AspNetCore.Diagnostics.Abstractions" />
</ItemGroup>

<ItemGroup>
<!-- this can be here, or can be done all at runtime in the Dockerfile -->
<!-- <PackageReference Include="ILLink.Tasks" Version="0.1.5-preview-1841731" /> -->
<PackageReference Include="Microsoft.AspNetCore.App" />
</ItemGroup>

</Project>

My strategy for figuring out which assemblies to "root" and exclude from trimming was literally to just iterate. Build, trim, test, add an assembly by reading the error message, and repeat.

This sample ASP.NET Core app will deploy cleanly on Zeit with the smallest image footprint as possible. https://github.com/shanselman/superzeit

Next I'll try an actual Microservice (as opposed to a complete website, which is what this is) and see how small I can get that. Such fun!

UPDATE: This technique works with "dotnet new webapi" as well and is about 73 megs per "docker images" and it's 34 megs when sent and squished through Zeit's "now" CLI.

Small services!

Sponsor: Rider 2018.2 is here! Publishing to IIS, Docker support in the debugger, built-in spell checking, MacBook Touch Bar support, full C# 7.3 support, advanced Unity support, and more.

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook bluesky subscribe
About   Newsletter
Hosting By
Hosted on Linux using .NET in an Azure App Service

How do you use System.Drawing in .NET Core?

September 12, 2018 Comment on this post [9] Posted in DotNetCore | Open Source
Sponsored By

I've been doing .NET image processing since the beginning. In fact I wrote about it over 13 years ago on this blog when I talked about Compositing two images into one from the ASP.NET Server Side and in it I used System.Drawing to do the work. For over a decade folks using System.Drawing were just using it as a thin wrapper over GDI (Graphics Device Interface) which were very old Win32 (Windows) unmanaged drawing APIs. We use them because they work fine.

.NET Conf: Join us this week! September 12-14, 2018 for .NET Conf! It's a FREE, 3 day virtual developer event co-organized by the .NET Community and Microsoft. Watch all the sessions here. Join a virtual attendee party after the last session ends on Day 1 where you can win prizes! Check out the schedule here and attend a local event in your area organized by .NET community influencers all over the world.

DotNetBotFor a while there was a package called CoreCompat.System.Drawing that was a .NET Core port of a Mono version of System.Drawing.

However, since then Microsoft has released System.Drawing.Common to provide access to GDI+ graphics functionality cross-platform.

There is a lot of existing code - mine included - that makes assumptions that .NET would only ever run on Windows. Using System.Drawing was one of those things. The "Windows Compatibility Pack" is a package meant for developers that need to port existing .NET Framework code to .NET Core. Some of the APIs remain Windows only but others will allow you to take existing code and make it cross-platform with a minimum of trouble.

Here's a super simple app that resizes a PNG to 128x128. However, it's a .NET Core app and it runs in both Windows and Linux (Ubuntu!)

using System;
using System.Drawing;
using System.Drawing.Drawing2D;
using System.Drawing.Imaging;
using System.IO;

namespace imageresize
{
class Program
{
static void Main(string[] args)
{
int width = 128;
int height = 128;
var file = args[0];
Console.WriteLine($"Loading {file}");
using(FileStream pngStream = new FileStream(args[0],FileMode.Open, FileAccess.Read))
using(var image = new Bitmap(pngStream))
{
var resized = new Bitmap(width, height);
using (var graphics = Graphics.FromImage(resized))
{
graphics.CompositingQuality = CompositingQuality.HighSpeed;
graphics.InterpolationMode = InterpolationMode.HighQualityBicubic;
graphics.CompositingMode = CompositingMode.SourceCopy;
graphics.DrawImage(image, 0, 0, width, height);
resized.Save($"resized-{file}", ImageFormat.Png);
Console.WriteLine($"Saving resized-{file} thumbnail");
}
}
}
}
}

Here it is running on Ubuntu:

Resizing Images on Ubuntu

NOTE that on Ubuntu (and other Linuxes) you may need to install some native dependencies as System.Drawing sits on top of native libraries

sudo apt install libc6-dev 
sudo apt install libgdiplus

There's lots of great options for image processing on .NET Core now! It's important to understand that this System.Drawing layer is great for existing System.Drawing code, but you probably shouldn't write NEW image management code with it. Instead, consider one of the great other open source options.

  • ImageSharp - A cross-platform library for the processing of image files; written in C#
    • Compared to System.Drawing ImageSharp has been able to develop something much more flexible, easier to code against, and much, much less prone to memory leaks. Gone are system-wide process-locks; ImageSharp images are thread-safe and fully supported in web environments.

Here's how you'd resize something with ImageSharp:

using (Image<Rgba32> image = Image.Load("foo.jpg"))
{
image.Mutate(x => x
.Resize(image.Width / 2, image.Height / 2)
.Grayscale());
image.Save("bar.jpg"); // Automatic encoder selected based on extension.
}
  • Magick.NET -A .NET library on top of ImageMagick
  • SkiaSharp - A .NET wrapper on top of Google's cross-platform Skia library

It's awesome that there are so many choices with .NET Core now!


Sponsor: Rider 2018.2 is here! Publishing to IIS, Docker support in the debugger, built-in spell checking, MacBook Touch Bar support, full C# 7.3 support, advanced Unity support, and more.

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook bluesky subscribe
About   Newsletter
Hosting By
Hosted on Linux using .NET in an Azure App Service

The Extremely Promising State of Diabetes Technology in 2018

September 07, 2018 Comment on this post [6] Posted in Diabetes
Sponsored By

This blog post is an update to these two Diabetes Technology blog posts:

You might also enjoy this video of the talk I gave at WebStock 2018 on Solving Diabetes with an Open Source Artificial Pancreas*.

First, let me tell you that insulin is too expensive in the US.

Between 2002 and 2013, the price of insulin jumped, with the typical cost for patients increasing from about $40 a vial to $130.

Open Source Artificial Pancreas on iPhoneFor some of the newer insulins like the ones I use, I pay as much as $296 a bottle. I have a Health Savings Plan so this is often out of pocket until I hit the limit for the year.

People in America are rationing insulin. This is demonstrable fact. I've personally mailed extra insulin to folks in need. I've meet young people who lost their insurance at age 26 and have had to skip shots to save vials of insulin.

This is a problem, but on the technology side there's some extremely promising work happening, and it's we have really hit our stride in the last ten years.

I wrote the first Glucose Management system for the PalmPilot in 1998 called GlucoPilot and provided on the go in-depth analysis for the first time. The first thing that struck me was that the PalmPilot and the Blood Sugar Meter were the same size. Why did I need two devices with batteries, screens, buttons and a CPU? Why so many devices?

I've been told every year the a Diabetes Breakthrough is coming "in five years." It's been 25 years.

In 2001 I went on a trip across the country with my wife, an insulin pump and 8 PDAs (personal digital assistants, the "iPhones" of the time) and tried to manage my diabetes using all the latest wireless technology...this was the latest stuff 17 years ago. I had just moved from injections to an insulin pump. Even now in 2018 Insulin Pumps are expensive, mostly proprietary, and super expensive. In fact, many folks use insulin pumps in the states use out of warranty pumps purchased on Craigslist.

Fast forward to 2018 and I've been using an Open Source Artificial Pancreas for two years.

The results speak for themselves. While I do have bad sugars sometimes, and I do struggle, if you look at my blood work my HA1c (the long term measurement of "how I'm doing" shows non-diabetic levels. To be clear - I'm fully and completely Type 1 diabetic, I produce zero insulin of my own. I take between 40 and 50 Units of insulin every day, and have for the last 25 years...but I will likely die of old age.

This is significant. Why? Because historically diabetics die of diabetes. While we wait (or more accurately, #WeAreNotWaiting) for a biological/medical solution to Type 1 diabetes, the DIY (Do It Yourself) community is just doing it ourselves.

Building on open hardware, open software, and reverse-engineered protocols for proprietary hardware, the online diabetes community literally has their choice of open source pancreases in 2018! Who would have imagined it. You can choose your algorithms, your phone, your pump, your continuous glucose meter.

Today, in 2018, you can literally change the code and recompile a personal branch of your own pancreas.

Watch my 2010 YouTube video "I am Diabetic" as I walk you through the medical hardware (pumps, needles, tubes, wires) in managing diabetes day to day. Then watch my 2018 talk on Solving Diabetes with an Open Source Artificial Pancreas*.

I believe that every diabetic should be offered a pump, a continuous glucose meter, and trained on some kind of artificial pancreas. A cloud based reporting system has also been a joy. My wife and family can see my sugar in real time when I'm away. My wife has even called me overseas to wake me up when I was in a bad sugar situation.

Artificial Pancreas generations

As the closed-hardware and closed-software medical companies work towards their own artificial pancreases, the open source community feel those companies would better serve us by opening up their protocols, using standard Bluetooth ISO profiles and use security best practices.

Looking at the table above, the open source community is squarely in #4 and moving quickly into #5. But why did we have to do this ourselves? We got tired of waiting.

All in all, through open software and hardware, I can tell you that my life is SO MUCH BETTER than it was when I was first diagnosed. I figure we'll have this all figured out in about five years, right? ;)

THANK YOU!

MORE DIABETES READING

* Yes there are some analogies, stretched metaphors, and oversimplifications in this talk. This talk is an introduction to the space to the normally-sugared. If you are a diabetes expert you might watch and say...eh...ya, I mean, it kind of works like that. Please take the talk in in the thoughtful spirit it was intended.


Sponsor: Get home early, eat supper on time and coach your kids in soccer. Moving workloads to Azure just got easy with Azure NetApp Files. Sign up to Preview Azure NetApp Files!

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook bluesky subscribe
About   Newsletter
Hosting By
Hosted on Linux using .NET in an Azure App Service

Always Be Closing...Pull Requests

September 05, 2018 Comment on this post [10] Posted in Open Source
Sponsored By

Always be closingI was looking at a Well Known Open Source Project on GitHub today. It had like 978 Pull Requests. A "PR" means "hey here's some code I did for your project, you can PULL it from here and merge it into your code!"

But these were Open Pull Requests. Pending. Limbo Pull Requests. Dating back to 2015.

Why do Pull Requests stay open?

Why do projects keep Pull Requests open? What's a reasonable amount of time? Here's a few thoughts.

  • PR as Call to Action
    • PRs are a shout. They are HERE IS SOME CODE and they create work for the maintainer. They are needy things and require review and merging, but even worse, sometimes manual merging. Plus for folks new to Git and Open Source, asking them to "rebase on top of latest" may be enough for them to just give up.
  • Fear of Closing
    • If you close a PR without merging it, it's a rejection. It's a statement that this work isn't going to be used, and there's always a chance that the person who did the work will feel pretty bad about it.
  • Abandoned
    • Sometimes the originator of the PR disappears. The PR is effectively abandoned. These should be closed after a time.
  • Opened so long they can't be merged
    • The problem with PRs that are open for long is that they become impossible to merge. The cost of understanding whether they are still relevant plus resolving the merge conflicts might be higher than the value of the PR itself.
  • Incorrectly created
    • A PR originator may intent to change a single word (misspelling) but their PR changes CRs to LFs or Tabs to Spaces, it's a hassle.
  • Formatting
    • It's generally considered poor form to send a PR out of the blue where one just ran a linter or formatter. If the project wanted that done they'd ask for it.
  • Totally not aligned with Roadmap
    • If a PR shows up without context or communication, it may not be aligned with the direction of the project.
  • Surprise PR
    • Unfortunately some PRs show up out of the blue with major changes, file moves, or no context. If a PR wasn't asked for, or if a PR wasn't requested, or borne of an Issue, you'll likely have trouble pushing it through.

Thanks to Jon and Immo for their thoughts on this (likely incomplete) list. Jess Frazelle has a great post on "The Art of Closing" that I just found, and it includes a glorious gif from Glengarry Glen Ross where Always Be Closing comes from (warning, clip has dated and offensive language).

Jess suggests a few ways to Always Be Closing.

Two things that can help make your open source project successful AND stay tidy!

What do you think? Why do PRs stay open?


Sponsor: Get home early, eat supper on time and coach your kids in soccer. Moving workloads to Azure just got easy with Azure NetApp Files. Sign up to Preview Azure NetApp Files!

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook bluesky subscribe
About   Newsletter
Hosting By
Hosted on Linux using .NET in an Azure App Service

Interesting bugs - MSB3246: Resolved file has a bad image, no metadata, or is otherwise inaccessible. Image is too small.

August 31, 2018 Comment on this post [7] Posted in DotNetCore
Sponsored By

I got a very strange warning recently when building a .NET Core app with "dotnet build."

MSB3246: Resolved file has a bad image, no metadata, or is otherwise inaccessible. 
Image is too small.

Interesting Bug used under CC https://flic.kr/p/4SpmL6Eek! It's clear, in that something is "too small" but what? A file I guess? Maybe it's the wrong size?

The error code is MSB3246 which is nice and googleable/searchable but it was confusing because I couldn't figure our what file specifically. It just felt vague.

BUT!

I had recently been overclocking my machine (overly aggressively, gulp, about 40%) and had a very nasty hard reboot. As a result I had a few dozen files get orphaned - specifically the files were zero'ed out! Zero is small, right?

Turns out you can pass parameters over to MSBuild from "dotnet build" and see what MSBuild is doing internally. For example, you could

/fileLoggerParameters:verbosity=diagnostic

but that's long. So how about:

dotnet build /flp:v=diag

Cool. What deep logging do I see now?

Primary reference "deliberately.zero.bytes.dll". (TaskId:41)
13:36:52.397 1:7>C:\Program Files\dotnet\sdk\2.1.400\Microsoft.Common.CurrentVersion.targets(2110,5): warning MSB3246: Resolved file has a bad image, no metadata, or is otherwise inaccessible. Image is too small. [S:\work\zero-byte-ref\zero-byte-ref.csproj]
Resolved file path is "S:\work\zero-byte-ref\deliberately.zero.bytes.dll". (TaskId:41)
Reference found at search path location "{RawFileName}". (TaskId:41)

Now with "verbose" turned on I can see that one of the references is zero'ed out/corrupted/bad. I reinstalled .NET Core in my case and doubled checked all the DLLs/Assemblies that I was bringing in - I also ran chkdsk /f - and I was back in business!

I hope this helps someone who might stumble on error MSB3246 and wonder what's up.

Even better, thanks to Rainer Sigwald who filed a bug against MSBuild to update the error message to be more clear. In the future I'll be able to debug this without changing verbosity!


Sponsor: Preview the latest JetBrains Rider with its built-in spell checking, initial Blazor support, partial C# 7.3 support, enhanced debugger, C# Interactive, and a redesigned Solution Explorer.

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. He is a failed stand-up comic, a cornrower, and a book author.

facebook bluesky subscribe
About   Newsletter
Hosting By
Hosted on Linux using .NET in an Azure App Service

Disclaimer: The opinions expressed herein are my own personal opinions and do not represent my employer's view in any way.