Techno: MS Office Transformation Server for Alfresco

Why would you need a Transformation Server on Alfresco?

There can be multiple reasons, but the 2 most reasons we’ve heard are the following.

1. There are performance issues at the moment
2. We need high quality PDF transformations

In our case we had a customer who had both :).

The first issue: Performance

Within an Alfresco infrastructure where one has a SOLR server, a REPO Server and in some cases an extra SHARE Server that’s probably the max of servers the big Alfresco community customers will go.
Something like the following diagram taken from the Alfresco documentation.

If one still has ‘performance’ issues on Alfresco Community then almost everyone will think directly about clustering without examining why the single Repository Server can’t handle the load.

When you have a good monitoring tool which monitors Alfresco & the Server, then after an extensive search it can become clear that LibreOffice can be the main issue here. In our case it didn’t help to have a LibreOffice manager to kill/restart the process in case it got stuck or running multiple processes.
Huh? is this even possible, yes it is we’ve made module just for this purpose called LibreOffice/OpenOffice manager

So even having Alfresco Enterprise level subsystem JodConverter managing LibreOffice didn’t even help. So in this scenario it wouldn’t even help switching to Enterprise.
For the record we at Contezza have both Enterprise and Community customers, so there is no fixation to use Alfresco Community in all scenarios :).

Back to the use-case disabling the transformations completely did help!
You all know the following screen when Alfresco isn’t able to load a preview, so it wasn’t long until this wasn’t acceptable any more by the customer.

The second issue: High Quality Transformations

Actually the issue is probably “Bad Quality Transformations” using LibreOffice.
Sure upgrading to the latest Beta version of LibreOffice did help a bit, but still there were issues which weren’t resolved.
For previews it didn’t really matter if the transformation sucks, but for official documents like when Signing stuff in Alfresco and you need to create a perfect PDF, then it matters.

Mostly the issue was related to the Header & Footer in documents. In a lot of cases we were able to reduce the issues by re-creating the MS Word documents/templates again.

The one thing we couldn’t fix was creating a PDF document from a MSG file :).
Yes we could previews the plain text (which Alfresco does as well), but that wasn’t sufficient to transform.
Hence the need of MS Outlook to do the official Transformation to PDF to get real PDF based on the HTML template used within MS Outlook.

In above example it’s clearly seen that the built-in preview/pdf isn’t suitable for official documents or to archive the PDF’s in Alfresco Records Manager.

Solution: Transformation Server

So basically we setup an external Transformation Server for both issues Performance & High Quality PDF’s.
We didn’t ditch LibreOffice completely, cause in terms of generating a lot of PDF’s , LibreOffice is much faster and reliable than MS-Office.
There is no MS-Office service available to generate PDF’s, so we only use MS-Office when needed the High Quality PDF.

The following chart describes how our server has been setup

Some key points of the Transformation Server:

• Running an Alfresco Repository with no direct access
• Disabled Cache
• Has access to the Database & Contentstore
• Disabled Alfresco Scheduled Jobs
• Needs a full MS-Office desktop license, we’ve used MS-Office 2013 at the time
• LibreOffice Manager installed
• Windows Print to PDF driver is used

We started this project somewhere in February and it was live in March. So our Transformation Server has now been running great for about 5 months. We waited on letting this now to the world to be sure it can handle the load.

Above is the chart of our monitoring tool since the last month :). The red horizontal line is a restart and the light red period was the notification that the disk was 95% full.
So average the Alfresco repo is below 10% CPU with 40-70 concurrent users on Alfresco.

Before we had long spikes till 100% and maybe weekly restarts of Alfresco until we disabled the transformation completely.

Need info on setting this up? Contact us and we can sure work something out.