Topic on Project:Support desk

A few days ago I upgraded our MediaWiki installation from 1.13.2 to 1.26.0, and PHP from 5.3.10 to 5.6.14. No change to web server and OS (IIS on Windows Server 2008). No change to MySQL (v5.5.22).

 When going to a valid page, it generally works ok. When there is anything wrong with the URL (e.g. article name is misspelled - even wrong case), the page fails with a 500 server error. Nothing is written to the PHP log file. (I also can't use colons in the URL (such as for category links). I need to encode to %3A in order for that to work.)

 The crash does not happen somewhere mid-way through the MediaWiki code. If it did, I might be able to debug and pinpoint where the problem is. I have confirmed that every line called by index.php finishes and then the server returns 500. It is as though the MediaWiki code inserts something into the output buffer that causes the PHP interpreter to crash as it is processing the output buffer.

 I can't provide the URL, sorry. This is a wiki in a private network and not accessible on the Internet.

 Does anyone have any ideas for how to troubleshoot this?

An error 500 is usually happening in the webserver. That means you should find more information about this error in the webserver error log.

What does it say?

I have been trying to post entries from the web server logs but I keep getting a message saying "This edit has been identified as spam" and to contact an administrator. How to contact an administrator is not clear so I will try to explain.

When I request a page that does not exist, the web server log file shows 2 entries: one for title=Special%3ARunJobs and one for title=does_not_exist. Both requests have 500.0 response codes.

I am also capturing the failed request tracing logs. They also agree that the 500 error was triggered by a request to: Special%3ARunJobs.

I had never heard of Special:RunJobs before having to troubleshoot these errors. Is it something that can be fixed or disabled?

The runJobs call is responsible for running jobs like the update of cached pages or so.

What you write sounds like you had a look into the webserver access log. What you however should investigate is, what the webserver error log is telling you about the error 500. It should contain more information about what is going wrong when these errors happen.

Again, when I try to post details from an error log file, I'm blocked because it looks like spam. I am invited to appeal to an Administrator, but the link doesn't take me to a page that makes it clear how to contact the Administrator. Any suggestions?

You can use the website "pastebin" or so to put some lines of text online.

Annoying, I know, but it should work.

Thanks for recommending pastebin. It works well.

You can see the output of the Failed Request Tracing at http://pastebin.com/raw.php?i=sM8dKUPS

The error is not meaningful to diagnose the problem and how to fix it.

Read Manual:Job queue to understand what's the purpose of the RunJobs URL. To mitigate the problem, you could set $wgRunJobsAsync to false, or set up a scheduled task to run jobs periodically and set $wgJobRunRate to 0.