Oracle WebCenter — Inheriting the Unmappable
I had never heard of Oracle WebCenter Portals or WebCenter Content before I joined this organisation. That turned out to be fairly normal. Most people who end up working with it seem to arrive the same way.
The platform is built on top of an Oracle database and sits inside highly controlled environment. My employer has been selling it as a packaged product for around twenty years — a collaborative working environment with serious access controls and document security. At the time it was built, the choice between WebCenter and SharePoint was a reasonable one. WebCenter was more mature, offered more, and the security model around portal visibility and document access was genuinely well suited to the use case. SharePoint has probably overtaken it since, but that’s not the point. The point is that twenty years ago this was a considered decision, and now it is my problem.
When I joined, I was told I would be looking after the platform. I asked how many portals we had. Nobody knew. I asked who the portal owners were. Nobody knew that either. What I did know was that there was a migration outstanding — old 11g-style portals needed bringing into 12c — and that I had no admin access, no documentation, and no predecessor to ask.
Getting admin access alone took weeks. Nobody had requested it before, the DBAs were not willing to give up control of the WebLogic account, and there was no documented process for what portal manager access even looked like. I got my development environment running first — an OVA file downloaded from Oracle, released in 2015 with no updates since, which took its own weeks of effort to get working — and used that to describe what access I actually needed. Once the service desk understood what a portal manager group was, they could add me to one. Small win. Took a while.
So how many portals did we have. Two hundred and forty-four.
The 11g versus 12c distinction turned out not to be two separate stacks, as I’d initially assumed. It was one stack, but some portals dated from before the 12c upgrade and carried the scars of it. Oracle had deprecated things like tab navigation and sub-portals in 12c, but the upgrade hadn’t removed the tabs — it had hard-coded them into the page using ADF. They existed, they just couldn’t be modified or removed. To change anything meaningful you had to build a new page. This explained a lot about why things were the way they were.
The platform has two persistent problems. The first is speed. The old portal strategy relied heavily on document list viewers — components that present the results of search terms — stacked across tabs, sometimes with tabs inside tabs. A single page load could be quietly calling hundreds of documents. Under any real load, the whole platform slows to a crawl. The second is fragility. There are error states — ADF Faces errors, mostly — that lock a session into a reload loop for hours. Session-based, so usually contained to one user, but the performance problem can ripple across everyone if two or three heavy pages get loaded at roughly the same time.
I spent the early months interviewing people, mapping what they said against what others said, and slowly building a picture of how things actually worked — portal creation, modification, user accounts, training requests, check-in forms. Nobody knew everything. Some people assumed others were responsible for things those others had never heard of. I drew it all out anyway.
Then a project landed. One division of the customer — around a thousand users — decided they wanted their portals modernised. My employer agreed. Now I needed to know which of the 244 portals belonged to this division, and the division wanted the portal owners to sign off on any changes. Neither we nor they had reliable ownership records. Tracking down the owners of 44 portals, while continuing everything else, took the best part of ten months.
The modernisation work itself has been more straightforward than the archaeology. The new template drops the tab-and-document-list-viewer approach entirely and uses elastic search to surface content from WebCenter Content instead. It’s faster, it’s cleaner, and most customers respond well to it once they understand that changing a portal doesn’t touch the content inside it. That explanation — content and presentation are not the same thing — is something I give several times a week, to customers and internal colleagues alike.
Along the way I’ve found efficiencies through trial and error and lab work, fixed things quietly where I could — a template that kept cloning an image with broken permissions, for example, got retired and replaced before it caused any more grief — and produced documentation for processes that had never been written down, including a standard operating procedure for portal retirement, because without one, portals just accumulated indefinitely.
I’ve been to Oracle. I’ve attended best practice sessions and asked specific questions that got specific answers. I was asked twice in about three weeks whether I’d take the title of Subject Matter Expert for the platform. I’ve declined both times — there is still too much I don’t know — but senior managers have started coming to me for updates anyway, including on migration work to 14c that isn’t formally mine.
I’m becoming the person who knows this thing. I didn’t plan for that. I’m not entirely sure how I feel about it. But the pattern of inheriting an undocumented system, making sense of it through persistence and lateral thinking, and quietly becoming more useful than the job description suggested — that part, at least, feels familiar.