We currently deploy web applications by creating a database and running SQL scripts through query analyzer. Then we copy the output from "publish website" and set up that website in IIS.
We have seen websetup in visual studio, but that part seems to be thinly documented. For example, we are not clear how to ask the user for IP and password of SQL server. We also tend to get websites deployed this way coming up under folders like http://example.com/project, instead of just http://example.com.
Then there are issues with AJAX.Net not being installed or some or the other patch not applied.
So far, we have physical access to the servers. Pretty soon though we are going to be shipping CDROMs. What is the practical tradeoff between manual intervention and automation?
Avoid Visual Studio deployment, and automate as much as possible. Web Deployment Projects and NAnt can be your friends!
Briefly, our deployment setup:
We use RedGate SQL to script differences between dev and live database.
An NAnt build file which calls MSBUILD to build the web deployment project (.wdproj), zips up the resulting compiled web app (along with the SQL change script) and then uploads the zip file to the server.
On the server side, there is another NAnt build file which takes the application offline, backs up the database, backs up the website. runs the SQL change script, unzips the new version and brings the app online.
Step 3 is usually run "manually" (one double-click), but sometimes scheduled for late at night. You could do exactly the same from a CDROM, or even write a pretty little Windows Forms app as a wrapper.
Quite happy to give details of the NAnt script if you're interested.
Have you tried using Web Deployment project? There is support for VS 2008 also now..
I deploy mostly ASP.NET apps to Linux servers. Here is my standard workflow:
I use a source code repository (like Subversion)
On the server, I have a bash script that does the following:
Checks out the latest code
Does a build (creates the DLLs)
Filters the files down to the essentials (removes code files for example)
Backs up the database
Deploys the files to the web server in a directory named with the current date
Updates the database if a new schema is included in the deployment
Makes the new installation the default one so it will be served with the next hit
Checkout is done with the command-line version of Subversion and building is done with xbuild (msbuild work-alike from the Mono project). Most of the magic is done in ReleaseIt.
On my dev server I essentially have continuous integration but on the production side I actually SSH into the server and initiate the deployment manually by running the script. My script is cleverly called 'deploy' so that is what I type at the bash prompt. I am very creative. Not.
In production, I have to type 'deploy' twice: once to check-out, build, and deploy to a dated directory and once to make that directory the default instance. Since the directories are dated, I can revert to any previous deployment simply by typing 'deploy' from within the relevant directory.
Initial deployment takes a couple of minutes and reversion to a prior version takes a few seconds.
It has been a nice solution for me and relies only on the three command-line utilities (svn, xbuild, and releaseit), the DB client, SSH, and Bash.
I really need to update the copy of ReleaseIt on CodePlex sometime:
http://releaseit.codeplex.com/
Related
I am stuck with a problem that I could use some feedback on to solve it in the best possible way.
The issue revolves around source control -> automated builds -> deployment. Basically ALM (Application Lifecycle Management).
We have a product – an ASP.NET Web application with a MS SQL database. This product is running on hundreds of websites with associated databases across multiple virtual machines in our production environment. At the moment the web applications and database are running on servers with IIS 7 and SQL Database Server 2008 R2. The product itself is source controlled in Team Foundation 2012.
For years the release of new versions of the product has been once or twice a year for years. Now we are going to focus on releasing more frequently and hence we need a strategy for the ALM for the product.
The deployment strategy now:
In the development period between the releases, the SQL update scripts has been created manually – each time a database change was made a script was updated. When the application is ready to be deployed it gets compiled on a developer machine. The database with all the changes used would be backed up into a .BAK file. The web application, the .BAK file and the update SQL script would be packaged (.zip) and uploaded to the production environment ready for deployment.
Update existing running products:
Copy/paste the web application in the target website physical folder.
Update the web.config file – connectionstring and application
variables. Run the update script via SQL Management Studio
This would be done for each and every customer – hundreds of times.
This is a very tedious and error prone task and I don’t like it at all!
What I would like to do instead is;
Source control the database as a Database Project in Team Foundation
Automatically build the web application with Team Foundation 2012
Build Server.
Deploy the output from the Build Server to the multiple websites of
the production environment along with automatically generated SQL
update scripts run against the SQL Server.
I have been googling my ass off - only finding bits and pieces regarding builds, deployment, automatic SQL update script etc.
What I think is partly the right direction is to source control the database and use the TFS Build Server. I am very confused on how to do the deployment itself in an easy and controlled way using the output from the TFS Build server.
Ideally I would want to the TFS Build server to create a package with the latest version of the Web application, the latest version of the database, post deployment script including an auto generated SQL Update script from the previous build to the current build. This could be contained in e.g. a nuget package. Then I would want to be able to create an additional web application which should manage the deployment – target, version, iis website, sql server, web.config connectionstrings etc.
Does anyone have any advice on how to achieve this? How do you do this?
You can use a release management tool to do this, no need to create an additional web application.
One such example is Deployment Manager, from Red Gate. (Disclaimer: I work there.) It has built-in deployment actions for ASP.NET apps and SQL Server databases. The command line tool RgPublish.exe can be used to create a package for the web app as you describe from TFS Build. The same can be done for the database using the sqlCI.exe command line and associated NANT/MSBuild scripts.
The same packages can then be deployed to each of your servers. You may run into scalability issues with 100s of websites though.
The database deployment works by generating the upgrade script automatically, though you can change the behaviour to put the upgrade script in the package, when the package is first built. These are called "dynamic" and "static" upgrade methods respectively.
I've been struggling while trying to create an automated TFS 2010 team build that uses MSDeploy to deploy the website to one of our local boxes. Aside from not being able to find good documentation, a large part of the problem is that the site uses a native C++ dll. I've managed to get things almost working now and the site will deploy during the build BUT the native dll is typically in use and prevents the deployment if the site has been accessed since the last restart. I would like to have the build force an application pool recycle or restart the site/application just prior to deploying the files to the remote server. I can use the following at the command prompt to achieve this manually:
msdeploy -verb:sync -source:recycleApp -dest:recycleApp="Default Web Site/Application",computerName="MyServer"
How do I hook this into the build process so that it occurs before the publishing pipeline tries to copy files to the server?
You could use the InvokeProcess activity in your workflow. You can then invoke a batch script that would contain your command line above.
You can drag the activity and insert it as the step before publishing (assuming you are using the workflow designer).
If you need the script to run as another user there are plenty of helpful suggestions here and in the linked blog.
I already owe SO and Ewald for getting our TFS build up and running.
The problem is, that when I commit project directory, there is uploaded everything including source code.
Not really sure why you want to upload via FTP? You shouldn't commit your own compiled binaries to source control for deployment though.
You could take a look at AppHarbor, just push your code with git and it will be build and deployed automatically.
more about AppHarbor
Real alternatives to Windows Azure PaaS (web role)?
Does it matter? Since asp.net pages can be compiled on the server, having source files on the web server is sometimes normal so IIS knows not to allow access to them.
That said, uploading output binaries into source control is generally a bad idea - it is better to do the deployment from your build server.
Actually, this is kind of hard.
For months, I've tried to automatize our deployment, without absolute success. For my experience, I can see only way to do that:
Have a build server on your deployment machine (or same network)
A build server will pull out your code from repository, say, once per minute and will check for modifications. If there's modifications, it will execute the build scripts related to this project. I suggest you to use TeamCity, because it is very easy to use compared to CruiseControl (I'm not sure if you can use Git with TFS). You can program your build server for build your solution or project and after, you can execute an msbuild script to copy the files to the production folder (e.g: c:\inetpub\yourapp or \\my_server\inetpub\yourapp). You can use MSBuild's Copy Task to do that.
UPDATE 1: I didn't tried, but if helps, you can push to an FTP server using git-ftp
UPDATE 2: Seems that some guy did some workarounds and successfully deployed his app using git and FTP.
Currently we deploy compiled ASP.Net applications by publishing the web site locally and emailing a zip file to the system administrator with a (usually) lengthy set of instructions for deployment. This is because the first time we deployed an ASP.Net application to a customer the dev and test IIS instance were the same, and we were unable to deploy the site twice to the same machine. This set the tone for deployment on all subsequent projects.
I am now evaluating our deployment methods and am looking specifically at the built in deployment tools; specifically I'm looking at custom installation tasks and using as much of the standard installer functionality as I can (mostly the user interface).
Secondly, I'm looking at merging deployments and automatic updates.
How do you go about deploying sofware in your organisation? What tools do you use, and what problems do you come across most frequently?
We have dedicated DEV, TEST, STAGE, and PRODUCTION servers.
We also have a dedicated build machine which runs Cruise Control.
Cruise Control is configured for a Continuous Integration build, which runs after code is checked in. It is also configured for separate Development, QA, Stage, and Production tasks.
To deploy to development, the code is first retrieved from SVN and built, then the "Precompiled Web" folder is copied to the development web site, and the web service project is copied to the development application server. Cruise Control is also configured to "tag" the source code before the build starts so we can reproduce the build at a later time, or branch from the tag if we need to do a hot fix.
To deploy to QA, the files are copied from the development machines to the QA machines.
Likewise, to deploy to Stage the files are copied from the QA machines to the Stage machines.
Finally, to deploy to production, the files are again copied from the Stage machines to the Production machines.
To configure each environment, we have a custom tool which is part of each environment's Cruise Control task that modifies connection strings, "debug=true|false", "customErrors=Off|RemoteOnly", and other environment-specific settings.
So each environment can be deployed with a button push from the Cruise Control dashboard.
One caveat is that we currently have the production database password configured in the Cruise Control config file...it would be nice move it elsewhere!
Lastly, let me add that even though our production machines are in a dedicated hosting facility, the servers are accessible from our Cruise Control machine, which makes it very easy to do a production deployment. The only manual step is to encrypt the web.config files and remove the "AppOffline.html" file that Cruise Control puts up.
Let me know if this helps, or if you have any questions.
Thanks!
A couple things that I have done is the following:
1) Use a Web Deployment Project in order to compile and clean the build as well as handing web.config section replacement if the config changes between environments.
2) Use NAnt to do all of the building, archiving, and copying in a repetitive manner.
The Web Deployment Project ends up creating a MSBuild file which can be used in place of NAnt; however, I came from a Java background and used Ant all of the time so NAnt is my preference in .Net. If you add in the NAnt Contrib tasks, you will be able to deploy not only the files but also handle items such as your source control (incase it is not part of the default tasks) and Sql Script Execution for changes.
Currently I use both of the options together. I have my NAnt build file call the Web Deployment Project through MSBuild. With the configuration manager setup for each environment, it allows me to manage the web.config section replacements automatically and still have fairly decent control over my copying and archiving of a release.
Hope this helps.
We use web deployment projects, and the VS 2008 projects to create an .msi from the output of the webdeployment & other projects. A normal windows app called 'setup' is used to do a lot of the db creation and preliminary stuff, rather than trying to customise the setup projects with custom steps. It is a lot easier to do this yourself than trying to customise the MS code. This windows app then calls the correct .msi files that the user needs.
Team foundation build runs every evening to rebuild the solution and copy everything to a 'Release CD' directory which anyone can access and do testing on the latest 'release'.
To be honest TFS build is a bit overboard for a small team like ours, and I only use it because its what I am used to.
In a previous company we used this http://www.finalbuilder.com/ and I can recommend it for ease of use and for the amount of software supported.
1) Build project with MSBUILD
2) FTP files to Production Environment
3) Copy / Paste manually to each web server
For intranet sites, we use CruiseControl in conjunction with SVN to have the site rebuilt automagically.
Theoretically you could extend this model over a VPN if you could map a drive remotely to a client's intranet. Or a more quick and dirty solution might be to use a tool like SyncBack to sync the remote folder containing the compiled DLLs for the site.
Deploy Web Applications Using the Copy Web Tool
Text from Microsoft Training Kit Book Web Based Development
Web Setup Projects are useful if you are providing a Web application to many users (for example, allowing people to download the application from the Web and install it). If you are responsible for updating a specific Web site for your organization, it’s impractical to log on to the Web server and install a Windows Installer package each time you make an update. For internal applications, you can edit the Web application directly on the Web server. However, changes you make are immediately implemented in your production Web application, and this includes any bugs that might be there. To enable yourself to test a Web application, you can edit a local copy of the Web application on your computer and publish changes to the production Web server using the Copy Web tool. You can also use the Copy Web tool to publish changes from a staging server to a production Web server, or between any two Web servers. The Copy Web tool can copy individual files or an entire Web site to or from a source Web site and a remote Web site. You can also choose to synchronize files, which involves copying only changed files and detecting possible versioning conflicts in which the same file on both the source and remote site have been separately edited. The Copy Web tool cannot merge changes within a single file; only complete files can be copied.
Does Microsoft offer a tool where you can deploy a web application to multiple web servers in a load-balanced environment/web farm?
My team is looking for a tool, preferably from Microsoft, where we can deploy our web application from development environment to production environment automatically.
If I understanding what your asking for your looking for a build server, to my knowledge Microsoft don't offer one, but some to take a look at are Team City, Hudson(requires a plug-in), and CruiseControl.net.
Basically they work by pulling from your source control building your application and running your tests. They all support scripting that will allow you to build then deploy to your servers. This can be set up to run nightly, weekly, etc. you can also set it up to monitor your source control for changes and build anytime it sees a change
The only one I've used is Team City, the install was easy, and depending on how many build agents you need it's free.
If your just looking to build and deploy from VS Another option is creating an NAnt script and running it from VS as an external tool.
For a good over view of Build servers check out this SOF question cruisecontrol.net vs teamcity for continuous integration
The Web Deployment Team blog at Microsoft has some reasonably useful information, and have a deployment tool you could try...
In the last environment we setup we used TeamCity for all our builds and deployments (Which is basically to say we wrote MSBuild scripts and automated them with TeamCity). In short we had the following 5 build configurations:
Continuous Build - Automatically rebuilt our product upon every check-in. Running all the tests. This build did not deploy anywhere
Nightly Build (Dev) - Automatically build and deployed our product to the development web server (no server farm). We build would run the tests, update the development database, shutdown the Dev IIS web site, copy the necessary files to our web server, and restart the site
Test Build - Like our Nightly build only it deployed to our test environment and it wasn't scheduled so it had to be manually started by logging into Team City and pressing a button
Stage Build - Like test only deployed to a web server that was externally visible to our customers sot that they could validate the application. Also, only run on demand.
Production - Created a zip file of our product that the deployment team could install on our production web servers
So I guess what I'm suggesting is that you use TeamCity and then write build scripts in such a way that they'll deploy to your Web Farm. If you want examples I could supply you with the pertinent portions of our build scripts
** One more thing: we check in our web.config files and such for each environment into subversion and then part of the build process is to copy and rename the appropriate config file for the environment. For example, web.prod.config => web.config in our production build
I believe that Sharepoint does this.
File Replication Service ( e.g. DFS Replication ) is a typical and very good choice for doing this.
Your changes are synced between member servers at the file system level.
Sharepoint does this automatically when you deploy a solution package.