Archive for the ‘Technology’ Category

Optional Kubernetes resources and PodPresets

Thursday, July 27th, 2017

The sample for Microservice Builder is intended to run on top of the Microservice Builder fabric and also to utilize the ELK sample. As such, the Kubernetes configuration for the microservice pods all bind to a set of resources (secrets and config-maps) created by the Helm charts for the fabric and ELK sample. The slightly annoying thing is that the sample would work perfectly well without these (you just wouldn’t get any logging to the ELK stack) only, as we shall see in a moment, deployment fails if the fabric and ELK sample have not already been deployed. In this post we’ll explore a few possibilities as to how these resources could be made optional.

I’m going to assume a minikube environment here and we’re going to try to deploy just one of the microservices as follows:

If you then perform a kubectl describe  for the pod that is created you’ll see that it fails to start as it can’t bind the volume mounts:

Elsewhere in the output though you’ll see a clue to our first plan of attack:

Doesn’t that optional  flag look promising?! As of Kubernetes 1.7 (and thanks to my one-time colleague Michael Fraenkel) we can mark our usage of secrets and config-maps as optional. Our revised pod spec would now look as follows:

And lo and behold, with that liberal sprinkling of optional attributes we can now successfully deploy the service without either the fabric or ELK sample. Success! But why stop there? All of this is boilerplate that is repeated across all our microservices. Wouldn’t it be better if it simply wasn’t there in the pod spec and we just added it when it was needed? Another new resource type in Kubernetes 1.7 comes to our rescue: the PodPreset. A pod preset allows us to inject just this kind of configuration at deployment time to pods that match a given selector.

We can now slim our deployment down to the bare minimum that we want to have in our basic environment:

Note that we have also added that runtime: liberty  label to the pod which is what we’re going to use to match on. In our advanced environment, we don’t want to be adding the resources to every pod in the environment, in particular,  we don’t want to add it to those that aren’t even running Liberty. This slimmed down deployment works just fine, in the same way that the optional version did.

Now, what do we have to do to get all of that configuration back in an environment where we do have the fabric and ELK sample deployed? Well, we define it in a pod preset as follows:

Note that the selector is matching on the label that we defined in the pod spec earlier. Now, pod presets are currently applied by something in Kubernetes called admission control and, because they are still alpha, minikube doesn’t enable the admission controller for PodPresets by default. We can enable it as follows:

(Note that, prior to minikube v0.21.0 this property was called apiserver.GenericServerRunOptions.AdmissionControl, a change that cost me half an hour of my life I’ll never get back!)

With the fabric, ELK sample and pod preset deployed, we now find that our pod regains its volume mounts when deployed courtesy of the admission controller:

Pod presets are tailor-made for this sort of scenario where we want to inject secrets and config maps but even they don’t go far enough for something like Istio where we want to inject a whole new container into the pod (the Envoy proxy) at deployment time. Admission controllers in general also have their limitations in that they have to be compiled into the API server and, as we’ve seen, they have to be specified when the API server starts up. If you need something a whole lot more dynamic that take a look at the newly introduced initializers.

One last option for those who aren’t yet on Kubernetes 1.7. We’re in the process of moving our generated microservices to use Helm and in a Helm chart template you can make configuration optional. For example, we might define a logging  option in our values.yaml with a default value of disabled , and then we can define constructs along the following lines in our pod spec:

Then all we’ve got to do when we’re deploying to our environment with the fabric and ELK sample in place is to specify an extra --set logging=enabled  on our helm install. Unlike the pod preset, this does mean that the logic is repeated in the Helm chart for every microservice but it certainly wins on the portability stakes.

Private repositories on Docker Hub

Sunday, July 16th, 2017

Sometimes Docker Hub really is just the quickest and easiest way to share an image from one place to another, particularly when the place I’m trying to share to is expecting to just do a docker pull. It’s not always the case that I want to share those images with the rest of the world though. Docker Hub’s answer to this is the private repository but, on a free plan, you only get one private repository. What you have to remember though is that a repository can contain multiple images: they all share the same name but each has a different tag.

So, a while back I created a repository in my personal namespace called private and made it private using the button on the settings page:

When I then want to push an image up I use the local name as the tag. For example:

Simple as that. There are obviously limitations here in that I lose the ability to have multiple versions of my image with different tags but so far, for my limited use cases, I’ve been able to live with that. In fairness to Docker Inc, I should say that have multiple private repositories is not the only reason to pay for an account on Docker Hub. You also get the ability to run parallel builds on.

Microservice Builder GA Update

Wednesday, July 12th, 2017

As I posted here on the Microservice Builder beta, I thought it only fair that I should offer an update now that it is Generally Available. There is already the official announcement, various coverage in the press including ZDNet and ADT, a post from my new General Manager Denis Kennelly, and, indeed, my own post on the official blog, so I thought I’d focus on what has changed from a technical standpoint since the beta.

If I start with the developer CLI, the most significant change here is that you no longer need a Bluemix login. Indeed, if you aren’t logged in, you’ll no longer be prompted for potentially irrelevant information such as the sub-domain on Bluemix where you want the application to run. Note, however, that the CLI is still using back-end services out in the cloud to generate the projects so you’ll still need internet connectivity when performing a bx dev create.

Moving on to the next part of the end-to-end flow: the Jenkins-based CI/CD pipeline, the Helm chart for this has been modified extensively. It is now based on the community chart which, most significantly, means that it is using the Kubernetes plugin for Jenkins. This results in the use of separate containers for each of the build steps (with Maven for the app build, Docker for the image build, and kubectl for the deploy) and those containers are spun up dynamically as part of a Kubernetes pod representing the Jenkins slave when required.

The Jenkinsfile has also been refactored to make extensive use of a Jenkins library. As you’ll see in the sample projects, this means that the generated Jenkinsfile is now very sparse:

I could say much more about the work we’ve done with the pipeline but to do so would be stealing the thunder from one of my colleagues who I know is penning an article on this subject.

Looking at the runtime portion, what we deploy for the Microservice Builder fabric has changed significantly. We had a fair amount of heartache as we considered security concerns in the inter-component communication. This led us to move the ELK stack and configuration for the Liberty logstash feature out into a sample. This capability will return although likely in a slightly different form. The fabric did gain a Zipkin server for collation and display of Open Tracing data. Again, the security concerns hit home here and, for now, the server is not persisting data and the dashboard is only accessible via kubectl port-forward.

Another significant change, and one of the reasons I didn’t post this immediately, was that a week after we GA’d, IBM Spectrum Conductor for Containers morphed into IBM Cloud private. In the 1.2 release, this is largely a rebranding exercise but there’s certainly a lot more to come in this space. Most immediately for Microservice Builder, it means that you no longer need to add our Helm repository as it will be there in the App Center out of the box. It also meant a lot of search and replace for me in our Knowledge Center!

You may be wondering where we are heading next with Microservice Builder. As always, unfortunately I can’t disclose future product plans. What I can do is highlight existing activity that is happening externally. For example, if you look at the Google Group for the MicroProfile community, you will see activity ramping up there and proposals for a number of new components. Several of the Microservice Builder announcements also refer to the Istio service mesh project on which IBM is collaborating with Google. It’s still early days there but the project is moving fast and you can take a look at some of the exciting features on the roadmap.

Multi-Stage Docker Build

Friday, May 5th, 2017

Docker 17.05 enabled the ability to perform multiple build stages in a single Dockerfile, copying files between them. This brings to regular Docker build a capability that I’ve previously talked about in the context of Rocker, and something that’s of particular use in a compiled language like Java. Let’s see what it would look like in the context of the WebSphere Liberty ferret sample.

The original Dockerfile looks as follows:

We can see that it assumes that the application has already been built and just pulls in the WAR file, in this case from Maven Central. With a multi-stage build we can perform the build of the application and the build of the image in a single Dockerfile:

The first line uses the Maven on-build image to build the application using the source in the same directory as the Dockerfile. Although the stages are given an index by default, naming them using the AS keyword makes the file much more readable. Further down in the Dockerfile we can see that the COPY command takes the built WAR file from the first stage and copies it into the Liberty dropins directory as before. The important thing about all of this is that the final image doesn’t end up with the application source in it, or Maven, or an SDK – just the artifacts that are needed at runtime – thereby keeping down the final image size.

Introducing Microservice Builder

Monday, March 27th, 2017

When the frequency of blog posts drops on this site it generally has two causes: I’m busy and/or I’m working on something that’s IBM Confidential. Both of these have been true over the past six months or so whilst I’ve been working on something we’re calling Microservice Builder. A public beta was announced in the run up to InterConnect which went live on the 24th which means that I can now come up for air and say a little about the work we’ve done so far.

Although not limited to Java deployments, Microservice Builder pulls together multiple strands of work that we’ve been doing in the WebSphere space. First, there is the work that is being done in the MicroProfile community to define a set of standard APIs for building microservices in Java. Initially, this took a set of existing Java EE technologies (JAX-RS, CDI and JSON-P) but now additional APIs are being defined. You can start to see the results of this work in the Liberty March beta where there are new features for injecting environmental configuration and utilizing fault tolerance patterns such as timeout, bulkhead and circuit breaker.

Another area where we’ve sought to improve the developer experience is by providing a fast-path to creating new projects. The Liberty App Accelerator has been around for some time now, allowing you to generate Java projects quickly through a web UI. We’ve taken this idea and extended it to cover Swift and Node.js. This can be achieved either through a web UI or through a new plugin to the Bluemix CLI. (Note that generated projects do not need to be deployed to Bluemix.) The plugin goes beyond just generating projects and allows you to build and run them locally using containers. This means that the developer no longer needs to have the prerequisites (e.g. Java, Maven and Liberty) installed locally.

For a runtime environment, we believe containers are a good fit for microservices and in the first instance we’re focusing on Kubernetes. That could be the newly announced Kubernetes in IBM Containers or it could be on-premises with IBM Spectrum Conductor for Containers. On top of Kubernetes, Microservice Builder adds a lightweight fabric, installed as a Helm chart, that simplifies deployment of Liberty-based services. Specifically, in this first release it generates key and trust stores to facilitate inter-service communication. It also configures an ELK (Elasticsearch-Logstash-Kibana) stack to receive and display information including trace, FFDC, garbage collection and HTTP access logs from the Liberty logstashCollector-1.0 feature.

The final strand of Microservice Builder ties together the development and runtime environments via a Jenkins based pipeline. Once again, this is installed as a Helm chart, and is configured to automatically pick up projects from a GitHub or GitHub Enterprise organization. For a Java application, the pipeline will build and test using Maven, before creating a Docker image and pushing it to a registry. The Docker image is then deployed to a Kubernetes cluster using either the same or a separate pipeline.

To show all of this in action, we have taken the sample conference application from the MicroProfile community and broken it apart in to separate projects to deploy using Microservice Builder. Just follow the docs to recreate it in either your local minikube environment or with Spectrum Conductor for Containers.

Presentations from IBM InterConnect 2017

Sunday, March 26th, 2017

I’m finally back home after what feels like a very long week in Las Vegas at IBM’s InterConnect conference. I promised that I’d post my presentations on SlideShare and I’ll add a few comments here on how each session went.

After an Inner Circle session on Sunday, my first public session of the week was an introduction to containers with WebSphere traditional. This played to a full room which suggests that there is significant interest in the use of containers for existing workloads. Indeed, that was the point of the second half of the session, to describe scenarios where it may make sense to use containers with traditional WebSphere. That’s not to say that it always does and, during one-to-one sessions during the week, I found myself repeatedly cautioning customers against rushing in to the use of containers, particularly with ND, just for the sake of it.

How to Containerize WebSphere Application Server Traditional, and Why You Might Want To from David Currie

My next session covered our new announcement around Microservice Builder. I’ll not say more here as I’ll cover this in a separate post.

Microservice Builder: A Microservice DevOps Pipeline for Rapid Delivery and Promotion from David Currie

Unfortunately, I didn’t get to deliver this session on Liberty and IBM Containers as it clashed with another that I was presenting. As touched on briefly in this presentation, one of the other announcements at the conference was for support for Kubernetes in IBM Containers. There was lots of excitement around this and I urge you to go and check it out for yourself.

WebSphere Liberty and IBM Containers: The Perfect Combination for Java Microservices from David Currie

On Wednesday I had a joint presentation with Brian Paskin looking at options for scalability with Liberty and containers. This was very much Brian’s presentation though so I shan’t post it here. There was an accompanying lab in the afternoon that looked at Liberty collectives and at IBM Containers.

My last session of the week was looking at some of the options when choosing a container orchestration platform: from Liberty collectives, through Swarm and Docker Datacenter, and Kubernetes with IBM Spectrum Conductor for Containers and IBM Containers. Many customers I spoke to this week were looking for a single definitive answer here but my response for now is still very much “it depends”.

Choosing a Container Platform for your WebSphere Applications from David Currie

Find me at IBM InterConnect 2017

Saturday, March 18th, 2017

I’m going to be at IBM’s InterConnect conference this coming week. If you’re going to be there too, there’s a quick run-down of the sessions I’ll be presenting below. The astute will notice that, due to a scheduling snafu, I’m supposed to be presenting two sessions at the same time on Tuesday. If you go to the Liberty/ IBM Containers session then I’m afraid you’ll have to make do with Tom – be kind to him!

If you want to chat about any combination of microservices, containers and WebSphere, you can find me on the microservices ped in the WebSphere area of the expo hall from 5-7:30pm on Tuesday and again from 3-5pm on Wednesday. I’ll be kicking off the latter with a live demo of Microservices Builder, of which more in another post. For Inner Circle customers, I’ll also be talking about this topic at 11am on Sunday.

HAJ-5451 : How to Containerize WebSphere Application Server Traditional, and Why You Might Want To
Date/Time : Mon, 20-Mar, 11:15 AM-12:00 PM
Location : Mandalay Bay South, Level 2 – Surf D
Presenter(s) : David Currie, IBM

BMC-7014 : Roundtable Discussion on Building Java Microservices with WebSphere Liberty
Date/Time : Mon, 20-Mar, 02:00 PM-02:45 PM
Location : Mandalay Bay North, Level 0 – Tropics A
Presenter(s) : Alasdair Nottingham, IBM; David Currie, IBM

BMC-7085 : Meet the Expert on IBM WebSphere Application Server Liberty on Docker
Date/Time : Tue, 21-Mar, 02:30 PM-03:15 PM
Location : Concourse, Bayside B, Level 1 – Meet the Experts Forum # 1
Presenter(s) : David Currie, IBM; Tom Banks, IBM

HAM-5526 : IBM Microservice Builder: A Microservice DevOps Pipeline for Rapid Delivery and Promotion
Date/Time : Tue, 21-Mar, 03:45 PM-04:30 PM
Location : Mandalay Bay North, Level 0 – Islander F
Presenter(s) : David Currie, IBM; Jeremy Hughes, IBM

BMC-5983 : WebSphere Liberty and IBM Containers: The Perfect Combination for Java Microservices
Date/Time : Tue, 21-Mar, 03:45 PM-04:30 PM
Location : Mandalay Bay North, Level 0 – South Pacific A
Presenter(s) : David Currie, IBM; Tom Banks, IBM

BMC-7014 : Roundtable Discussion on Building Java Microservices with WebSphere Liberty
Date/Time : Wed, 22-Mar, 08:00 AM-08:45 AM
Location : Mandalay Bay North, Level 0 – Tropics A
Presenter(s) : Alasdair Nottingham, IBM; David Currie, IBM

BMC-2714 : Utilizing WebSphere Application Server Liberty in Docker Containers for Scalability
Date/Time : Wed, 22-Mar, 10:15 AM-11:00 AM
Location : Mandalay Bay North, Level 0 – South Pacific A
Presenter(s) : Brian S. Paskin, IBM; David Currie, IBM

HAJ-2718 : Utilizing IBM WebSphere Liberty in Docker Containers for Scalability (Lab)
Date/Time : Wed, 22-Mar, 01:00 PM-02:45 PM
Location : Mandalay Bay South, Level 3 – South Seas H
Presenter(s) : Brian S. Paskin, IBM; David Currie, IBM

BAS-5901 : Choosing a Container Platform for Your WebSphere Applications
Date/Time : Thu, 23-Mar, 10:30 AM-11:15 AM
Location : Mandalay Bay North, Level 0 – South Pacific A
Presenter(s) : David Currie, IBM; Tom Banks, IBM

How many processors?

Tuesday, January 31st, 2017

Reading Daniel Bryant’s O’Reilly publication Containerizing Continuous Delivery in Java reminded me of the challenge of determining how processors are available to you when running in a container. In the case of Java, a call to Runtime.getRuntime().availableProcessors() should show this all important information. A quick check reveals that, when called in an unconstrained container, this correctly returns the number of cores on my physical hardware (Docker on Linux) or assigned to the VM containing the Docker Engine (Docker Toolbox or Docker for Windows/Mac). If I used the --cpuset-cpus option on docker run to constrain the cores available to the container then this is also correctly reflected in the value returned. The difficulty arises when access to those CPUs is constrained in other ways.

Take, for example, the new --cpus option in Docker 1.13. Setting this to two on a four-way box, I still get four back from a call to availableProcessors() and rightly so: there are four processors and I may get simultaneous access to all four of them even if the cgroup is then going to make sure that I don’t get that access for more than half of the time. Another potential constraint is a highly multi-tenant environment. If I deploy my test application to Bluemix it tells me that there are 48 processors. That’s great but I’m pretty sure I’m not going to get exclusive access to all of those!

One example we’ve seen where this becomes a real problem is in native memory usage. By default, WebSphere Liberty uses the number of available processors to decide the number of parallel threads it should support. Each of those executors utilises a thread and each of those threads takes up space in native memory. In a containerized environment where total memory is typically constrained (Bluemix containers are sold by the GB/hour) and some generic heuristic is often used to determine the heap size to allocate to the JVM, that can lead to memory exhaustion. That’s why you’ll see a GitHub issue from my colleague Erin that, among other things, proposes setting hard-coding a maximum on the number of threads for the executor service in our Docker images.