Monday, 29 October 2007

STSADM export, Content Deployment, Content Migration API, Features/Solutions - deployment options compared

Back in May I wrote a post titled SharePoint deployment options : Features or Content Deployment?, which discussed some thoughts on what was the "right" way to move assets from development to production (and perhaps environments in between) during the site development process in SharePoint. Having now worked on other projects and consciously used different deployment methods on each, I'm rapidly coming to the conclusion that the "right" way to do deployment varies according to circumstances. So I thought what might be useful is an analysis of the whole range of deployment options, with information which might help you decide more easily on how you will complete this crucial step of the process.

So let's run through the options and their characteristics. Note by the way, that none of the options use 'destructive synchronization', where all content is first deleted before import.

Using STSADM export/import

Description:

Uses STSADM commands to generate a file (export) which can then be transferred to the target for import. One of the simplest ways of moving content from one place to another, although unlikely to be suitable as a continuous deployment mechanism. Examples:

stsadm.exe -o export -url http://localhost -filename C:\Export.cab -includeusersecurity -versions 4 -overwrite

stsadm.exe -o import -url http://localhost/sites/newsite -filename C:\Export.cab -includeusersecurity

Good for:

  • Moving an entire site/web as a one off
  • Quick deployment tests
  • Reparenting webs (can be into a different site collection)

Considerations:

  • Content on target will be overwritten if already exists
  • Granularity down to web only
  • Object GUIDs are not preserved (so some things will need to be 'fixed up' e.g. anything that references a list by GUID - ListViewWebPart, using lists with InfoPath forms)
  • Not a backup/restore tool - although it's the option which is most like backup/restore, things like alerts, audit trail, recycle bin items, security state, workflow tasks/state are not exported
  • Not transactional

Using Content Deployment via Central Admin *

Description:

Configured via 'Content Deployment paths and jobs' in Central Admin ('/_admin/deployment.aspx'). A path defines the source/target and authentication details, specific jobs define exactly which content should be deployed and how often. Quick deploy functionality allows users with permissions to specify important content which should be deployed more regularly than existing job schedules configured by administrators (quick deploy items are deployed every 15 mins).

Good for:

  • Moving entire site collections/webs on a scheduled basis e.g. in an authoring/production or authoring/staging/production topology
  • Deploying only incremental changes, e-mail notifications of success/failures
  • Allowing site owners to have some control over content deployment via Quick Deploy
  • Automatically deploying dependencies of content selected for deployment, even if in different site (e.g. page layouts/content types/site columns/referenced images etc.)
  • Automatically transferring the deployment package to the target environment (via HTTP[S])
  • Not transactional

Considerations:

  • Content on target will be overwritten if already exists
  • Granularity down to web only
  • No differentiation between site content (e.g. pages/images) and site 'infrastructure' (e.g. master pages, page layouts)
  • Object GUIDs are preserved
  • Blank site template should be used for source and destination site collection (see http://support.microsoft.com/kb/923592)
  • Also not a backup/restore tool (see above)

Using the content migration API *

Involves writing code which uses the content migration API (known as PRIME) to export then import content - the API is easy to use.

Good for:

  • Complete flexibility over deployment options
  • Granular control over what gets deployed (down to item level)
  • Ability to preserve object GUIDs (so that list GUIDs do not need to fixed-up)
  • Ability to select options for security, versioning and user roles

Considerations:

  • Blank site template should be used for source and destination site collection (see http://support.microsoft.com/kb/923592)
  • Not transactional
  • Also not a backup/restore tool (see above)
  • Need development skills to write code

Using Features/Solutions

The focus of this blog for several articles. Involves defining XML configuration files which SharePoint uses to add artifacts in the correct way on the target. This can be significantly more complex than simply developing in SharePoint Designer but can allow for better management throughout a solution's lifecycle.

Good for:

  • Iterative development/deployment
  • Deployment of assemblies and filesystem files (none of the other methods deal with this)
  • Ability to deploy assemblies/filesystem files to all servers in a farm with Solution packages
  • Possibilities for continuous integration

Considerations:

  • Developer is responsible for evaluating and deploying dependencies (e.g. underlying content types).
  • Updates to content types, list definitions, site columns etc. deployed via a Feature must be done with the API - modifying original Feature files and then reprovisioning is not supported
  • Can be very time-consuming due to lack of assistance from current tools

* Some additional notes on using Content Deployment or content migration APIs:

- appropriate Features will automatically be activated on the target, but they must be present (i.e. installed) for content deployment to work (N.B. publishing feature should not be enabled on target for first deployment)

- using Content Deployment or content migration API with RetainObjectIdentity option should not be combined with STSADM -export/import, since the latter will allocate new IDs!

So clearly there can be a few aspects to consider in choosing how to go about deployment for your project. In many scenarios where Features/Solutions aren't the most appropriate option, I favor using the content migration API, mainly due to the flexibility which isn't provided in any of the other options. Of course it does mean writing code, but as I mentioned last time, I'll soon share the mini-app I wrote so you don't have to!

Some useful references:

Sunday, 14 October 2007

Deployment using STSADM export or content migration API

Having focused on deployment using Features for several articles, back in May I wrote an article titled SharePoint deployment options : Features or Content Deployment?, which explored some of the decisions around deployment strategies for SharePoint projects. There are a variety of methods which can be used to move SharePoint artifacts and content from one place to another, and I think it's fair to say there's still a certain amount of confusion around deployment for many SharePoint developers. I certainly wouldn't claim to have all the answers, but after delivering another project last week, it seems like a good time to go over some of the experiences and reflect on the different approaches.

Needless to say, as far as deployment strategies go in general, the best idea is to have one! I see many newsgroup posts from people approaching the end of the development phase asking how they should move their work to the live servers. The problem I find with leaving deployment until the end of the project, is that none of the approaches are completely straightforward (particularly depending on what your solution consists of), and so if your project is to be delivered on time, it's important to know what steps you might need to go through.

As a sidenote, let's clarify some potentially confusing terminology here:

  • Content Deployment - the "paths and jobs" functionality which can be used to move content, surfaced by screens in Central Admin
  • Content Migration API - the underlying API (sometimes referred to as PRIME) which actually is used for both STSADM export and Content Deployment (in slightly different ways), in addition to the 'Manage Content and Structure' tool and in migrations from CMS2002

This time round I had decided to use the content migration API to deploy our solution, and it worked well for our circumstances. This is a contrast to developing with Features which I've done in the past, and the main reasons for choosing this approach were: 

  • no need for iterative deployment - although our overall project is phased, for this component we were able to develop the solution and then deploy everything from our development environment. (This approach will not work for subsequent deployments since content our client has generated on the live site would be overwritten on each deployment - more on this in an upcoming post.)
  • ability to retain object GUIDs - this simplified deployment significantly for our project, since if our lists were allocated new GUIDs on deployment (as happens with STSADM export/import), our components which referenced these lists (ListViewWebPart, InfoPath forms etc.) would not hook up properly on the deployment target. This would add a lot of "fix-up" steps to the deployment process if we were to use STSADM export.
  • no direct HTTP access from source Central Admin to target Central Admin - this is a prerequisite to use the Content Deployment functionality (paths and jobs) in Central Admin, but what we needed was a file we could copy to the live server. The content migration API provides this ability and also gives a compression option for large amounts of data.
  • automatic inclusion of database dependencies - as with STSADM export, (but not with Features), SharePoint will analyze and collect all dependencies such as fields, content types, master pages etc. for us.

The API is fairly simple to use and you may have seen Stefan's series of excellent articles on the subject - these serve as a good companion to the MSDN documentation.

It's important to remember that any non-database assets (e.g. user controls, assemblies etc.) need to be deployed manually to the target environment - these will not be included by use of something like STSADM export or the content migration API. In our case, since the live environment was a single server (and versioning would be handled by our main source control system), these were deployed by XCOPY since deployment using Solution packages did not offer any compelling advantages here. 

Whilst we're talking about filesystem files, it's useful to be aware that if you see 404 errors on the target after performing (e.g.) an STSADM import, chances are you've forgotten to deploy something like a user control. The 404 is actually coming from the referenced file rather than the actual page, so don't assume something has gone wrong with the import - a check on 'View all site content' and the import log will probably confirm all the site pages are present!

Hopefully this has given some food for thought on an approach you may not have considered. I guess my main message here is that whilst STSADM export is extremely simple, it may not provide the complete answer to all your deployment challenges due to changing GUIDs. In upcoming posts I'll provide a more direct comparison of deployment strategies (extending my 'Features or content deployment' post), and also share my mini-app which provides a front-end onto the content migration API.

[P.S. Sincere apologies to people who left comments whilst I was on holiday which are still not published - I'll publish these and respond over the next few days.

C.]