We've asked Atlassian support for help with this problem, but they haven't been able to give us any answer besides "we're investigating", so we thought we'd try the community to see if anyone else is having this problem:
We recently upgraded our Jira server to version 8.1.0, our Jira Service Desk plugin to 4.1.0, and our Confluence server to 6.13.4. After the upgrade, our service desk agents and customers have reported that the knowledge base articles returned during searching from the service desk portal were not relevant to the search, severely impacting the usefulness of the KB. After some investigation, we noticed that the issue seems to be related to spaces in the request. As an example, I've attached four screenshots of the issue we're seeing. In the first screenshot, I did a search for "what is an SBA". We have an article in our KB named exactly that title. However, none of the results were related to SBA at all. In the second screenshot, I did a search for simply SBA and the 3rd result was the article I was looking for. I also attached a screenshot of Chrome developer tools network tab. While the service desk portal page is correctly using the encoding %20 for spaces in the request (see the URL in the location bar), the AJAX REST requests are being sent to the backend using %2520 for the spaces. In the last screenshot, I just displayed the results for the REST service using the correct encoding.
We do have an Apache reverse proxy on our production servers, but we then tried installing "vanilla" versions of Jira and Confluence (straight out of the box installations, using their installers, no plugins) on both Windows and Linux and we *consistently* are seeing the same problem. Atlassian support has indicated they were not able to recreate the issue.
We then tried upgrading Jira to 8.2 and Service Desk to 4.2, and while the problem with spaces went away, the same problem is recreated when simply adding a ? to the end of the search query. Given that we use KCS for our knowledge base, many of our customers and agents add a question mark to the end of their searches, and so upgrading as a solution isn't a permanent fix for us. We also found that adding other special characters to the search (& ! and #, among others) had random effects on the search results, including erroring out.
It looks like the issue is with the encoding of the request between Jira Service Desk and the backend server, but we've tried a number of different things to fix it, and haven't found any that have worked.
Can anyone in the community do better than Atlassian support and please offer advice on how to fix this? This is severely impacting adoption of Jira Service Desk and the knowledge base at our organization. Thanks!