Docker workshop

Overview:

We recently carried out a short introductory Docker workshop, starting from scratch, installing Docker and taking it through to the point where a software stack, consisting of several linked containers, are deployed using docker-compose. Here’s what we covered.

Docker concepts:

Docker containers are easy-to-deploy units of software, analogous to  the shipping containers used by the transport industry, which simplifies the job of shipping diverse goods around the world.

Docker images are the templates for the containers. Every Docker container is started from an image. Images are defined by a Dockerfile which contains instructions for building the image, based on an existing image (for for instance, a web-server image will be based on an OS image, simply adding a layer of web-server software to it).

A Docker registry is where images are stored. Every machine where Docker is installed has a local registry. Additionally, Docker provides a central registry (from which images are fetched if they aren’t available locally). And finally, you can host your own private registries.

Starting point:

A freshly installed Ubuntu 16.04 server, called docker-test.

Installing Docker:

We won’t use the standard Docker package available from the Ubuntu repositories because Docker is changing fast – instead we’ll add the Docker apt-repository and install the latest version from there.

The last two commands ensure that we can run Docker commands without sudo.

Check that Docker is correctly installed with:

If everything is OK, Docker will download the hello-world image from the central Docker registry and run it.

Building a simple website container:

The purpose of Docker is to allow you to build and deploy something like a website without having to worry about details like which OS is to be used, which web-server etc. What we want is to have an image called say “website” which takes some files we give it and publishes them via a web-server.

First create a directory “website” where we will work on creating our container.

Create a text file called “Dockerfile” containing the following:

This defines a new image, based on the existing image “alpine”, a compact Linux OS image. A web-server, “lighttpd” is installed. Then our website content (index.html) is copied to the web-server content folder. and finally the web-server is started.

Next create a simple index.html as content for the website:

Now build an image from the Dockerfile and call it “website”.

The “docker images” command will show us the newly built image.

We can now run a container based on our new image as follows:

This will bring up the website container based on our “website” image and publishing the web content on port 8088 (the -p parameter maps the host port 8088 to the standard web-server port 80 within the container).

The “docker ps” command will show us the running container:

Point a browser at http://docker-test:8088 and you’ll see our simple “Hello World!” web-page, served by our new container.

Basic Docker commands:

Persistent data (stateful containers):

When a container is rebuilt from an image, it loses any changes which were made to it since it was last built. In order to preserve state (for instance, a database container will usually need to preserve its database contents, even if the container is rebuilt), this state must be maintained by the host and provided to the container by means of “volumes”.

To illustrate this, we’ll create a database container using the postgresql database server. First we create a data directory on the docker host, which will maintain the persistent state of the database.

Then we create a database container based on a standard “postgres” image from the main Docker registry. We pass the -v parameter instructing docker to map the host directory (~/data)  to the container path /var/lib/postgresql/data (where postgresql stores its database contents).

If you now look in the host directory ~/data, you will see that postgresql has created a set of database files there. Note: you’ll need to use sudo to list the files because postgresql has modified the file permissions.

Now connect to the new database server (docker-test:5432, user: postgres, password: postgres) with a postgres client (e.g. pgadmin) and create a database called “test”.

To demonstrate that the host is maintaining state for the container, we’ll now recreate the container and image from scratch.

Then we’ll run the container again (which will again fetch the image from the main Docker registry because we removed it from the local Docker registry with the “docker rmi” command).

Reconnect with postgres client – our new “test” database is still there even though we rebuilt the container (because we specified a persistent volume).

docker-compose – deploying a whole stack

The philosophy of a container is that its supposed to do just one thing well – this reduces complexity and increases reusability. So you shouldn’t use a single container to deploy several components. For instance, if you have a web application stack which consists of a database, a REST server, a client web application and a proxy server, then this stack should be deployed as four containers.

For this workshop, we’ll deploy a database and a REST server as a stack, using docker-compose to deploy the stack in a single operation.

We first need to install docker-compose (its an add-on tool for Docker).

Now we’ll create a small REST server in python and deploy it in a container.

Enter the following python code into test.py:

Our sample REST server will respond to a GET request to /hello with “Hello from the rest server!”

Next we’ll create a Dockerfile for the REST server image

And enter and save the following content into Dockerfile:

Now we can build and run the container.

Point a browser at http://docker-test:5000/hello and the REST server should return “Hello from the rest server!”

So now we have a REST server running as a container. The next step is to hook up the rest server to the database, so instead of always returning a fixed text string, it can do a more real-world task of returning the result of a query against the database.

Using the postgres client, connect to the database and create a database “test”, with a table “test” and one int column “test”. Insert a few values which our REST server will sum.

Now we’ll update the code of our REST server to sum the values from the test table.

We’ll need to update the REST server image to install the psycopg2 database library

Point a browser at http://docker-test:5000/hello and the REST server should now return “20” (the sum of the two values in the database).

So we now have two containers, one of which uses the other.

However, we are still starting both separately and in a specific order. We also currently have ports 5000 (rest server http) and 5432 (postgres) open and we have a hard-coded reference to “docker-test” in the rest server code. We could of course allow the database server to be passed in as a command line variable or as an environment variable, but docker-compose provides a better way, by linking the containers, so that the database is first started, then a private network is created to link the two containers and the address of the database server on that network is passed to the rest container, who can then communicate privately with its database container.

Lets update our REST server code, to reference the database as “database” instead of the host name “docker-test”. We’ll then use docker-compose to ensure that the host name “database” will be pointing to the database container.

Now we’ll create a docker-compose.yml file to link the REST server to the database server, by means of the host-name “database” (defined by the name of the service in the docker-compose.yml file). We no longer need to expose the postgresql port (5432), since docker-compose will provide a private network between the containers, allowing the REST server to access port 5432 inside the database container without it being exposed to the host.

The depends_on instruction ensures that the REST server container is started only after the database server container has been started.

Remove the current “rest” and “database” containers, rebuild the rest server image (with the updated test.py) and bring them back up as a stack with docker-compose

Check it again by browsing to http://docker-test:5000/hello (the REST server should still return “20”).

If you run the “docker ps” command, you’ll see that the containers are no longer called “rest” and “database”, but that docker-compose has constructed names based on the service names in the docker-compose.yml file.

You’ll also noticed that the database container is no longer publishing port 5432 – its only available on the private network which docker-compose creates between the containers. So this means that only port 5000 – the port published by the rest server – is now exposed to the outside world.

Note that we’re doing everything (building the image and deploying the container) on a single machine – in the real world, the docker images are built during the development cycle on a developer workstation (or a continuous integration server like jenkins) and pushed to a remote registry (like docker hub or a private registry). During deployment, the images are pulled from the remote registry and the containers started on the production server.

Docker swarm

Docker swarm is a very interesting extension to Docker which allows you to deploy your software stacks across a cluster of worker machines. Its easy to set up – one Docker host creates the swarm and becomes the swarm manager, and other hosts join the swarm. Using an extension to the docker-compose.yml, software stacks can be deployed across the cluster with replicas for fault tolerance and load balancing. We didn’t have time to cover docker swarm in this workshop, but we’ll cover it soon.

Contact us!

We’ve already gathered lot of experience using Docker to help our customers efficiently deploy their software stacks. If you’re interested in having us help your organisation get up and running with containers, get in touch with us at info@armstrongconsulting.com

clustered cron jobs

We’ve been recently running rest APIs on active-active server pairs (docker containers running on pairs of VMs on separate hosts) with postgres-BDR (multi-master bidirectional replication) for fault-tolerant storage and a pair of fault-tolerant HAProxy instances for incoming request routing. This is a robust setup which provides zero downtime during rolling updates or hardware maintenance or failure.

However, clustering scheduled jobs (i.e. ensuring that scheduled jobs execute exactly once) becomes a problem in this configuration. Multi-master replicated databases avoid a single point of failure but they are not suitable for use with database-based clustered schedulers like Quartz, so we needed to consider other options. There are complex clustered job schedulers, but we wanted to keep it simple and use Linux crond for scheduling. We finally settled on using keepalived to maintain a single master across the cluster and schedule the cron jobs identically on all servers, using a small if_master.sh script to ensure that a crontab entry only runs if the server is currently the keepalived master. The crontab entries then look like:

if_master.sh checks if the server is currently keepalived master (by looking for the floating ip address)

and only executing the command if the server is currently master (the same crontabs execute on the other servers, but do nothing if the server is not the keepalive master).

This was the simplest solution we could find which allows us to keep using cron and configure crontabs on all servers identically.

 

Finally, perfect wifi coverage at home

We live in an old house, on three levels. Its always been a challenge to achieve consistent wifi coverage throughout the house. We neglected to install ethernet cabling when we renovated and have been struggling with wifi issues ever since. We tried power-line networking (Devolo, TP-Link) and, although it worked most of the time, it provided very inconsistent performance and it was impossible to figure out why. We then reverted to a central wireless router and range extenders (Apple, TP-Link). Coverage was pretty bad in many parts of the house. Last weekend, we installed the AmplifiHD mesh networking system from Ubiquiti and we finally have the full performance of our internet provider (40-60Mbps LTE, depending on the time of day) from any or all devices, anywhere in the house.

516535-ubiquity-amplifi-hd-home-wi-fi-system

Replacing VMWare with KVM

I’ve been using an i5 Intel NUC at home as a home server.

I initially installed ESX on the NUC and ran an ubuntu VM with iptables, DNS, DHCP etc. However, I wanted to put the firewall between the home network and the LTE router, so I needed two network interfaces. The NUC only has one, so I thought I’d use VLANs to split the network.  That turned out to be pretty complicated to manage so I ended up buying a USB3 ethernet adapter (AX88179) for the NUC instead.

Getting that to work with ESX was a pain (I tried pass-through, but couldn’t get it to work reliably), so in the end I replaced ESX on the NUC with KVM. Worked great – the USB ethernet adapter installed out of the box with Ubuntu and the whole configuration only took an hour to set up.

Like with ESX, I still need an extra VM on my macbook (Ubuntu desktop  under Fusion) to run virt-manager to manage KVM (instead of the Windows VM where I used to run the vmware client to manage the ESX installation).

I used to backup the VMs with ghettoVCB to a Synology via nfs (which worked completely reliably). Now I’ll use LVM snapshots and and file copy to backup the KVM VMs to the same Synology.

Powerline networking with mixed Devolo and TP-LINK

We’ve been using Devolo powerline networking at home for years – we have an old house with wifi-proof walls. We’d been having plenty of problems with them, possibly due to overheating of the wifi plugs. After replacing the dLan 500 adapters with the more expensive Devolo dLan Pro adapters without any improvement, I finally decided to try TP-LINK AV500s instead.

The TP-LINKs are much cheaper (base plug plus two WLAN repeaters for < EUR 100). They are also compatible with the Devolo dLan 500s, so I’m still using a few of the Devolos (for a printer and in the cellar). Unfortunately they don’t perform better than the Devolos – I still get about 30Mbit throughput (measured with iperf between two devices connected by ethernet to the TP-LINKs). Should’ve installed CAT5 when wiring the house :(.

Addendum: the Devolos did not play so well with the TP-LINK network, so I ended up replacing them with some additional TP-LINKs. For the time being, the network seems to run reliably.

eCobertura makes code coverage easy

We used to wait for Jenkins to produce us a Cobertura report and of course nobody read it until delivery time arrived and we realised we hadn’t met the SLAs. Enter the eCobertura eclipse plugin which provides you with visual code coverage directly from within the eclipse editor. Just run your unit tests and you immediately see the source lines in green and red. Wow! How did we ever do without this?

Unable to start VMs on ESX5i

We came across a weird problem on ESX5i. Occasionally, one of our hosts would suddenly be unable to start any VMs – the running VMs were fine, but any attempt to start new ones would fail with an “Unknown internal error”. The first time this happened, I restarted the management agents and finally suspended all VMs and rebooted the server, after which everything was OK again for a few months and then the problem occurred again.

This time I decided to figure out what was going on. The log files in /var/log contain a lot of useful information and I was able to see that the problem was actually caused by ESX being out of disk space on the device used for /var/log. What was happening was that the driver for the Adaptec 5405z RAID controller in the machine was writing a huge log file which was not being rotated, so after a few months it consumed all the disk space.

The workaround was to add a line to the crontab (note: you also have to add a line to /etc/rc.local to readd the line to the crontab, otherwise it’ll be lost on next reboot) which deletes the adaptec log file periodically:

/etc/rc.local

echo "0 0 * * 0 rm /var/log/arcconf.log" >> /var/spool/cron/crontabs/root

/var/spool/cron/crontabs/root

0 0 * * 0 rm /var/log/arcconf.log

Seems strange that such a robust product as ESX5 doesn’t protect itself against this situation.

Anyway, the moral of this post is that if ESXi is producing any error messages you can’t interpret, have a look in /var/log.

Simple form validation in Wicket

I used to add custom FormValidators to forms where multi-field validation was required (like checking if two copies of an entered password match.

However, this approach has some problems – I had to list the dependent components and if some of them were hidden at validation time, the validator didn’t work.

Reading around, I saw the recommendation to use the onValidate() method of the form to do validation and this seems a lot more straightforward.

Here’s a sample wicket form with some validation logic to check if the current password matches and if the two copies of the new password are the same.

[code language=”java”]
final Form form = new Form("form", new CompoundPropertyModel(user)) {
private static final long serialVersionUID = 1L;

@Override
protected void onValidate() {
super.onValidate();

User user = model.getObject();
String current_password_entered = currentPasswordField.getConvertedInput();
String new_password_entered = newPasswordField.getConvertedInput();
String confirm_new_password_entered = confirmNewPasswordField.getConvertedInput();

if (current_password_entered != null
&& !User.encryptPassword(current_password_entered).equals(user.getPassword()))
error(getString("current_password_not_correct"));

if (new_password_entered != null && !new_password_entered.isEmpty()) {
if (new_password_entered.equals(current_password_entered)) {
error(getString("new_password_same_as_current_password"));
} else {
if (!new_password_entered.equals(confirm_new_password_entered))
error(getString("new_passwords_dont_match"));
}
}
}
};
add(form);
[/code]

Fault-tolerant ESX datastores for free

Preamble: NFS works great for ESX datastores. Its a whole lot easier to manage than iSCSI and although iSCSI is generally considered to perform better, we find that the flexibility of using NFS more than makes up for the lost performance.

If you have sufficient budget, there are great solutions from Dell, HP etc. where you can get fault-tolerant ESX installations already setup in a rack, which not only provide data-store fault tolerance, but also VM failover and so on. But there’s a lot of people or companies out there who have several servers running the free ESX5i hypervisor and who would still like to have some fault-tolerance. This article is for those people.

What’s not so great about shared storage (like NFS or iSCSI) is that you generally have a lot of extra boxes around and complicated network configuration. If you want fault tolerance, you’ll need two ESX servers, two NFS servers and two switches (and a bunch of cables to connect them all). Furthermore, you’ll add a lot of complexity to configure those (often proprietary) NFS boxes. I recently configured a couple of Lefthand boxes for a customer and it not trivial to set up.

So I figured there must be an easier way – after all, ESX is an amazing platform for reducing the number of boxes in the rack, so why would I want to start adding boxes again if I don’t have to.

The first important point is that ESX provides the VMXNET3 10Gb virtual ethernet adapter, so that even if your ESX server does not have 10Gb network cards, the VMs running on the server can communicate with each other at 10Gb and, more importantly for our purposes, ESX itself can communicate with its VMs at 10Gb speeds. So if we run an NFS server as a VM on the ESX server, and use it as an NFS datastore for the ESX server, then the server will see it as a 10Gbps NFS server.

OK, but what about the fault-tolerance? For that, we need to replicate the NFS server’s disk to another server. So, if we don’t have a real 10Gb network, that’ll have to be across 1Gb. Does that slow things down? Apparently not much – we’re using DRBD asynchronously which causes a minimal performance hit.

So what we do is to clone the NFS server VM to a second ESX server and set up DRBD replication between the two VMs. We’re using Ubuntu 11.10 server (you’ll need a reasonably recent Linux distribution to get the 10Gbps support with the VMXNET3 virtual network adapter).

Because you only get the 10Gbps datastore access to the NFS server when the NFS VM is hosted on the local ESX server, this is not really shared storage (or at least its shared only at 1Gbps to other ESX servers). However, for our fault-tolerance purposes that doesn’t matter much. In fact, from a scalability point of view, it makes sense to provide each ESX server with a locally-running NFS datastore, accessed at 10Gbps and replicated to another ESX server. This scheme also has the advantage that each ESX server is autonomous – ESX servers with remote datastores always make me a bit nervous – any problems on the network and the VMs are likely to freak out. This way, the ESX server is completely self-contained – it only needs another ESX server for fault-tolerance. Even if the network fails, the local NFS datastore will be unaffected (except that fault-tolerance is temporarily suspended) and when the network is available again, the DRBD secondary will simply catch up again automatically, providing fault-tolerance again.

ESX Server 2 can do the same thing with another pair of NFS servers (a local one for fast access and a remote one on ESX Server 1 for fault tolerance). This idea can be scaled indefinitely – each ESX server having its own local NFS VM running its datastore and replicating to another ESX server. The major advantage of this approach is that its more scalable than a single fault-tolerant pair of shared NFS servers and you get 10Gbps access for free. Conversely, the price you pay for this is that you have a separate NFS server for each ESX server which makes administration more complex than for a single shared datastore (but hey, you can’t have everything, at least not for free).

You could additionally configure the nfs servers to fail over a shared ip address to the secondary – we haven’t bothered to do this since if the primary nfs server fails, its likely that the whole ESX server has failed. If that’s the case, we need to promote the DRBD secondary to primary manually, start the NFS server and register the VMs.

And how does it perform? Pretty well actually. The benchmarks below are made on an ESX5 host with a single i7 930 CPU, 24GB RAM, an Adaptec 5405z controller and 4x SATAII disks in RAID5.

Disk performance of a VM running directly on the ESX host (i.e. on the local datastore).

hdparm -tT /dev/sda

/dev/sda:
Timing cached reads:   12396 MB in  2.00 seconds = 6201.81 MB/sec
Timing buffered disk reads: 398 MB in  3.00 seconds = 132.59 MB/sec

dd if=/dev/zero of=ddfile bs=8k count=20000
20000+0 records in
20000+0 records out
163840000 bytes (164 MB) copied, 0.274105 s, 598 MB/s

And now the disk performance of a VM running on our fault-tolerant nfs datastore:

hdparm -tT /dev/sda

/dev/sda:
Timing cached reads:   12242 MB in  2.00 seconds = 6124.43 MB/sec
Timing buffered disk reads: 258 MB in  3.00 seconds =  85.89 MB/sec

dd if=/dev/zero of=ddfile bs=8k count=20000
20000+0 records in
20000+0 records out
163840000 bytes (164 MB) copied, 0.707547 s, 232 MB/s

As you can see from the above, the VM running on the fault-tolerant NFS datastore is not as fast as the VM running on the local datastore, but it’s sufficiently fast for the subset of your VMs which require more fault-tolerance than that provided by daily backups.

Speaking of backup, we’re using ghettoVCB to backup 150 VMs (>1TB) every night via NFS to a Netgear ReadyNAS Ultra 6 device. We then rsync them to a remote data-center for offsite, versioned storage.