Submitted by ackbar on May 27th, 2004

Facts

On May 26, 2004, Slashdot underwent database maintenance, crippling the main functionality of the site. The maintenance started during primetime browsing hours, and lasted well into the night. As shown by this screenshot (showing the last few comments made, taken at approximately 9 PM CDT), the last posts before the outage were made at 8:29 PM CDT.

The nature of the maintenance allowed users to view Slashdot normally, but not to post comments or moderate. This screenshot was taken after trying to post a comment.

During the outage, a new story about ReactOS and Wine was posted by simoniker. This story went without comments for over four hours, with the previous Pixar story sitting at 14 comments throughout the entire outage.

The maintenance concluded at 12:40 AM, when users were again able to post and moderate. Load the Pixar thread to see the results of the outage- a handful of posts are made from 8:24 to 8:29 PM, then the timestamps jump to 12:40 AM. The first post to the ReactOS thread are shown in this screenshot.

However, all was not well- at 8:40 AM, functionality was still not back to normal. The front page was not reflecting the amount of posts made. It usually only takes a minute or so for this to happen, but at the time of this screenshot, these stories weren't brand new— one was 20 minutes old with about 15 posts, the other one hour old with approximately 80 comments.

At the time of this writing, the front page was still not reflecting the number of comments in a timely manner.

Commentary

This is yet another example of Slashdot's unprofessional production quality. The maintenance began during prime browsing hours- the time when those unable to read Slashdot during the work day would be doing their day's browsing, and lasted well into the night- the major browsing time of nocturnal nerds. Instead of having a short maintenance period at a time of low usage, an extremely lengthy update occurred at a time of high usage, which means that the update was probably unplanned.

The length of the maintenance also shows that it was unplanned. When planning database maintenance, the commands to be executed are determined prior to the maintenance, usually on a test server. These commands are executed in the shortest time possible once the site's functionality has been disabled. It is pretty clear that the admins didn't know exactly what needed to be done, given the lengthy outage. If the commands had been tested and planned, the maintenance would have been brief, and there would have been no residual effects such as the front page reflecting old comment counts.

The almost surely unplanned and lengthy outage once again demonstrates the lack of testing or version control on part of Slashdot's full-time development and administrative staff.

---

More injustices by other
Injustices Home