Sitecore Continuous Integration with Team City and TDS

CIProcess

There are a lot of articles around on how to do automated deployments / continuous integration with Sitecore, which if you’re new to the tools will likely leave you slightly baffled. This article will hopefully show you exactly what you need to do and explain why.

Solution Overview

  1. TDS is used by developers to serialize their Sitecore item changes and push them into source control
  2. Team City is used to detect the changes and run a build script
  3. Team City uses Web Deploy to push the code changes to the web server
  4. Team City calls MSBuild which will trigger TDS which is installed on the server to deploy Sitecore items to the destination server

Prerequisites

  • You have a build server with Team City installed and know how to set it up to do a web deploy
  • You are already using TDS and have your Sitecore items serialized in source control
  • Essentially you know how to do the first 3 steps and just need help with Step 4

Step by Step

UNC Share

On your web server you need to set up a UNC Share on your website’s folder. When TDS does a deploy it will install a web service on your website through this share, do the item deployment and then remove the web service again.

The share needs to give permission to the user that your Team City Build Agent runs as. To find out which user your Build Agent is using:

  1. open the list of services and find TeamCity Build Agent in the list
  2. right click and select “Properties”
  3. in the “Log On” tab you will be able to see which Windows User is being used

Team City Config

In your Team City’s build configuration settings for your project, add a new build step with the following config:

Runner Type: MSBuild
Step Name: Publish TDS Items (or some other identifier)
Build file path: Path to your projects .sln file
Command line parameters:

  • SitecoreDeployFolder: TDS will use this file path to install a web-service on your site to publish the items through.
  • SitecoreWebUrl: This is the url of the site you are going to update. TDS will use this when it tries to call the web service it installed.
  • IsDesktopBuild: false
  • GeneratePackage: false
  • RecursiveDeployAction: Delete
/p:IsDesktopBuild=false;GeneratePackage=false;RecursiveDeployAction=Delete;SitecoreWebUrl=URL OF SITE;SitecoreDeployFolder="UNC PATH TO YOUR SITECORE SITE"

Setting the command line parameters here will take precedence over any that have been included in your TDS projects solution file (which are liable to be overwritten by a developer).

TDS Build Settings

That’s it!

It’s that easy. If you run your build script now your items should all be published to Sitecore.

Alternatives

This certainly isn’t the only way to setup automated deployments and nor is it without issues. The fact a share needs to be set up between the Web Server and the Build Server, could cause an issue with security and may just not be possible if you’re using a cloud server.

Rather than using TDS to deploy the Sitecore items you could use TDS to create a .update package. These would normally be installed through an admin webpage (not great for CI) but there is an open source project called Sitecore Ship that will expose a REST endpoint for the package to be posted to. Brad Curtis has written an excellent guide to this setup here (http://www.bradcurtis.com/sitecore-automated-deployments-with-tds-web-deploy-and-sitecore-ship/), however at the time of writing Sitecore Ship isn’t compatible with Sitecore 7.5 or 8.

Another alternative to installing the update package is to use the TDS Package Installer. This is a tool Hedgehog provide alongside TDS for installing the update package. In this scenario you would need the tool installed on your web server and some way to call it. Jason Bert has written a setup guide for this example (http://www.jasonbert.com/2013/11/03/continuous-integration-deployment-with-sitecore/) however as well as Team City, you will also need Octopus Deploy to call the package installer. Octopus Deploy works by having what it calls Tentacles on each server you deploy to, making it easy to set up scripts to call programs on that server.

Sticking with the example using just TDS, you could also use TDS to deploy the solutions files as well as Sitecore items rather than using Web Deploy. However the downside here is that TDS is unable to modify your Web.Config file, which is one reason to stick with Web Deploy.

What’s your source control strategy?

I’ve seen companies that have no form of source control, companies using backups as source control, people still using Source Safe along with a whole bunch of actual good source control solutions such as Git and SVN. But when I say what’s your source control strategy I don’t mean what tool are you using, I mean how are you using it.

Checking you’re code into source control adds many benefits such as merging, versioning, reverting etc, but to get the most out of it you really need a strategy to define how you will use it. There is no one right way and the strategy you use ultimately depends on your team size and what you are doing.

Here are a few example strategies:

The didn’t know about a strategy approach

You’ve got a source control solution, you’re regularly checking your code in but you’ve never made a branch.

This is actually quite common. Everyone is developing against the same branch, you’ve got all the benefits of seeing who changed which piece of code, you can revert when you need to revert and when you do an update there’s some help with merging.

What you don’t get though is the ability to have work happening in parallel and one being released before the other. If you needed to do a release either you have to have a way of hiding the unfinished updates as the code is pushed to live or some way of reverting just those bits.

A branch for every feature

The other extreme is to have all development work happen on a new branch. When it’s ready to be release the code is merged back into the trunk and then released (plus some testing along the way).

The advantages of this are your trunk is always the same as what’s on live meaning emergency fix’s can be made without trying to find the last release in the source control history. Developers can also work separately without stepping each other toes and can release.

However with more than one development happening at the same time, whoever merges second could have a big job on their hands. All that branching is also going to be time consuming.

Updating branches often

So you’re using branches and to avoid a tricky merge are keeping the branch up to date regularly, possibly daily. Anyone who’s tried to merge a feature branch after 3 months of development can testify that it doesn’t always go smoothly and can take some time.

Only updating the branch at the end

You’ve decided to accept the large merge at the end of the project. You don’t like it but figure all those small merges during the project actually add up to more time than one big one at the end. You can then also concentrated on getting the feature built without the regular merge distraction.

A tag for each release

If you’ve ever needed to find a previous release and only have the history to go on it can be hard. If there’s no comment that says release to live you are basically guessing based on release date and last commit.

A tag is basically the same as a branch and if its part of your release plan, you’ve got a good way of finding each release to live.

A branch for live, a branch for dev

In this setup you have a live branch that always match’s what’s on the live site. There’s less need for tags as commits to this branch are normally also a release to live, and because the branch is always in sync with live, emergency fix’s are taken care of.

The downside with this approach is everyone is working on the dev branch, so exceptions should be made and feature branches used when it makes sense.

So…

So there are some of the strategies I’ve come across, but there will be plenty more. What you do ultimately relates to what kind of work you do, how frequently you release and what kind of team size you have. But the one thing you must do with all of them is make comments on each commit. If you don’t then source control becomes virtually useless.

IIS Where are my log files?

This is one of those things that once you know is very very simple, but finding out can be very very annoying.

IIS by default will store a log file for each site that it runs. This gives you valuable details on each request to the site that can help when errors are being reported by users.

When you go searching for them your initial thought may be to go to IIS and look at the site you want the files for. There you will see an item called logging. Excellent you think, this will tell you all you need to know.

IIS Log Files

There’s even a button saying “View Log File…”, but once you click it you realise things aren’t so simple. The link and folder path on the logging page both take you to a folder, containing more folders. In those folders are the logs, but there’s a folder for each site in IIS and they’ve all got a weird name. How do you know which folder has the log files for the site you want?

IIS Log Files Folders

Back on the IIS logging screen there’s nothing to say which folder the files will be in. There isn’t any indication anywhere.

The answer however is very easy. Each folder has the Site ID in its name. You can find the Site ID for your site in IIS either by looking at the sites list

IIS Sites List

or clicking on a site and clicking advanced settings

IIS Advanced Settings