Loading...

Issue with sporadic hang is fixed

In recent weeks and months we have experienced an issue whereby the web projects become unresponsive - basically unable to serve your requests. During this time we have kept a very close watch to make sure we get it back online as soon as possible. Still, we know that some of you no doubt have been inconvenienced by these short downtimes.

The good news is: We have discovered and fixed the cause of the issue.

A bug in [[TOC]] wiki macro only appeared when requests where made to a page in a project that used the macro, and such a project would then permanently claim one database connection. It 'infected' a project on first view, but it did not get worse by itself. However, at random intervals pages with the macro was requested on other projects tying up new connections until at some point no further database connections could be established.

(Those that are technically inclined can read about the issue and see my fix at http://trac-hacks.org/changeset/4366)

Finding this bug has taken much longer than we had imagined. However, it has led us to review all the central parts of the code, and fixing and reworking some parts that were not optimal. Additionally, we have updated all server software to the latest and greatest versions. The service is now in very good shape.

We can only apologize, and thank you all for being patient with us trough this period.

  • Posted: 2008-10-02 09:31 (Updated: 2008-10-03 07:59)
  • Author: simon
  • Categories: operations news

Comments

No comments.