Content down due to the database

Some of the more regular readers may have found out that there has been hickups with Nomadig.com content. I experienced this yesterday by myself when trying to manage the posts — I got a WordPress error message instead of the list of the posts.

WordPress documentation instructs to repair MySQL tables, but I ran check table and repair table commands for the main post table, but everything was ok from the MySQL’s point of view.

The problem went away suddenly and I shrugged and continued to hack something else…

This morning, the content was down again. Now I digged deeper, first with WordPress and then tried to run the same query in MySQL command line client, and got same error from the database. The nice folks at WordPress didn’t mention that the error code 28 is in reality from the database itself and not generated by WordPress.

This discovery led me to the MySQL site that instructed to check whether the temporary disk space is full. And there was the real problem. OnSmart fixed this fast and now Nomadig.com seems to be alive and kicking — hopefully for good.

It’s sort of sad (and also unacceptable) that the absence of space for temporary files can cause such drastic effects. I was merely reading data, not writing it. Of course, database engine needs to have some space for spilling out stuff from the memory, if the datasets grow too large. So at the end of the day, the problem is in WordPress that relies to the database in every operation.

Maybe I should investigate on having a disk based cache mechanism for the content. The delivery would be speedier and hopefully more robust — I know that there are problems with this approach, too, when Linux disks get full; files get truncated and so forth. But at least the majority of the pages would be available…