XCP = Xtra Complicated dePloyment

Greetings,

For this post, I am hoping to reach some ears within the Product Engineering department of EMC. Specifically, regarding the installation of the xCP stack.

During the first couple of releases it was very complicated and rather awkward to install all of the components to make an xCP system. That was expected as xCP really is just a combination of about 9 or so different products (give or take, depending on what you really need) that all require their own separate installations and configurations to work correctly. As the product aged with newer releases, I fully expected EMC to make it more like a true “product” as they tout it as such. Man, was I naive. I’ve been working specifically with Documentum for going on nine years now and have been primarily infrastructure focused. I’m also EMC|Proven to the specialist level with the EMCSysA cert among others and so I’m very familiar with Documentum and many other complicated products.

I guess since EMC was pushing xCP as a product with their product bundles and from catching wind of EMC wanting to basically rename Documentum to xCP, get everyone used to xCP being the best BPM/Workflow/Task-Driven enterprise software that does ECM great as well, and then once that happened, to rename xCP back to Documentum so that customers would instantly think of Documentum as the number one Business Process Suite. That’s how I heard it. Not sure if it’s true or not. Sounded like a decent marketing plan. The problem, however, is that xCP is still not a single unified “product,” rather, it’s a bunch of separate products that still are very much separate and as such all require their own setup and specific hardware/software requirements. Going through all of them individually is rather painful and one can easily get caught-out by some very small one-off requirement that can throw your design off and make you look like you don’t know what you’re doing.

This was the case just recently when the client decided they wanted to add the Brava viewer to their TaskSpace applications. And, while EMC doesn’t own Brava (it’s made by IGC), they did build the integration for it (but IGC still sells it), they should be able to drive some changes to make their software play better with others. What I’m referring to is the fact that the use of the Brava viewer (which, by license, only views TIFF and PDF files inline, office formats require another license) requires a dedicated Windows host for each environment (DEV, TST, QAT, PRD) to serve as a simple License Server. When you’re standing up completely Red Hat Linux environments and that requirement gets thrown-in at the last minute, it really throws a wrench in the works. What’s really odd, is that you can’t even get access to the Brava documentation until your client has paid for the licenses. In our case, the decision of Webtop vs. TaskSpace hadn’t been finalized until late in the builds and all four environments had been built without examining Brava specific prerequisites. Then comes along this simple viewer with it’s own specific and bloated requirements of four Windows servers, OS licenses and all. That’s really not what this blog is about, but it’s just case in point about the xCP product not really being a product at all, yet.

Let’s get on to what I’m blathering about. To get a basic xCP environment stood-up, the following products are required to be installed separately (not including desktop fat client software):
Content Server (Docbroker and Docbase)
Process Engine
DAR Files deployed to the Docbase: BPM (installed with PE, but often required to reinstall), Forms, TaskSpace,
TaskSpace
Documentum Administrator
Business Activity Monitor (BAM) (requires an App Server separate from TaskSpace and a database)

In our environments, we also had the following additional products installed:
xPlore Index Server
Brava!
MyDocumentum for Outlook
MyDocumentum Offline
Document Reporting Services
Crystal Reports Designer
Retention Policy Services (RPS)
Content Storage Services (CSS)

The majority of the above products require separate installations and many are installed on dedicated hosts. This makes the installation process somewhat complicated. The interesting part of xCP is really around the installation and configuration of the BAM product as it has several components and they have very specific configuration requirements including the Web App and the Database component. The biggest gripe is why there are so many config files to get it pointed at the right database, content server, docbase, etc. It seems to me this could be made much simpler by having all the pertinent configuration information placed in a single properties file which is read from to populate the various xml files. Also, in building four distinct environments, the database and repository population scripts never work correctly the first time, no matter the order they are run in. We tried both ways, several times and neither worked. You will have to run the cleanup-db and cleanup-repo scripts to clear out the docbase and BAM database. Oh, and that also has it’s own special configuration file, which contains the same info you had to enter in the main xml config files to get BAM to install in the first place. Why all these separate files? It leaves too much room for a fat finger to get in there and really hose things up. Then you spend your day looking for the issue which turns out to be an “m” instead of an “n” or something silly.

It just seems to me they could really clean up these installations, perhaps making a bigger installer that allows you to select the products you wish to install on the given host you run it from. The EMC SourceOne software works this way. Perhaps I’m being naive again and the expectation that the xCP “product” should install like a single “product” is dreaming too big? I don’t know. But I do know all of the ins-and-outs of all the products that are needed to make an xCP system, so that’s good, at least until the next release.

This post isn’t meant to slam EMC or to point fingers at issues with the products, I’m really just hoping that someone, someday, will recognize some of the silliness around the installation of xCP. Work it out gang. There’s always room for improvement, and getting the software installed and working, it seems to me, should be rather straight-forward and simple.

Hopefully, in the future, EMC will “Xcellerate” the Composition Platform installation.  I have heard that xCP 2.0 is supposed to address some of these gripes I have, but we wont know until its actually released.

-Scott Johnson
Infrastructure Architect

3 thoughts on “XCP = Xtra Complicated dePloyment

  1. Hello Scott – Your post has indeed found the ears (well, eyes) of a product manager at EMC. I handle the Cloud and Virtualization strategy for xCP, so hopefully I can lend some insight to our ongoing product activities.

    Anything I discuss here has already been presented in the public setting and subject to disclaimer (any forward-looking statements are subject to change, etc.).

    For the reasons you outline in the post, we focus xCP 2.0 on 1) consolidating several products into one, true product in order to radically improve the development experience, and 2) optimizing for virtual and cloud environments. Since the post is titled Xtra Complicated dePloyment, let me give a bit more detail around our plans for environment provisioning and application deployment.

    With xCP 2.0 we aim to reduce the time it takes to stand up production-quality environments from potentially months down to hours. We optimize for VMware’s vFabric application tier and use a policy-driven approach to automate the provisioning of Content Server, xPlore, application server (tcServer), BAM, and other elements of the Documentum platform and xCP stack. Our new provisioning engine, (very) tentatively called xCelerated Management System (xMS) uses blueprints that EMC will provide to automate the process of standing up and configuring clustered, highly available, xCP environments that adhere to the most rigorous industry best-practices. Also, once you stand up an environment using xMS it is wired out-of-the-box for various levels of monitoring via Hyperic…the monitoring solution provided as part of vFabric.

    With this approach, customers and partners can provision a complete environment in a fraction of the time it can take today..and standardize the way they stand up development, test, staging, and production systems going forward.

    That’s it for now. I just wanted to let you know that your concerns are being heard, and we are actively working on delivering a new, innovative approach toward cloud and virtual deployment with xCP 2.0.

    Like

  2. Great to hear it, RMan. Can’t wait to see how it comes out. With our involvement in the CAP program and working with that groups leaders as well as the lead Application Architect for EMC, hopefully we’ll see things early enough to actually provide valuable insight that can help EMC avert the patchset 13, patchset 14 etc challenges of late.

    So, can’t wait for it. Thanks for the reply.

    Like

  3. Scott,

    I am having the same issue but there is one thing I don’t understand. During the Brava for TaskSpace installation it wants to make changes to the TaskSpace files. How did you do that? Do you have a network drive on your windows server connected to the Linux TaskSpace server so you can point to the TaskSpace deployment?

    Thanks for any info,
    Ed

    Like

Leave a comment