SQL server was overloaded


Joren and Robinson made me aware of a problem with the wiki, ownCloud and other sites on the kermit server using SQL backends, including the main website. The SQL connections were all in use by the MediaWiki upgrade script that has been running for a few hours already over the wikihelp instance, which seem to have an _extremely_ large cache (a magnitude of the regular wiki cache).

Sorry for these issues - we will try to fine-tune MySQL again (already invested a fair amount of time in doing so before), and if that doesn't help, we probably need to invest in some more hardware.

We're growing heavily these days - which is good news, of course, but has even higher infra demands... to give you a number: SQL sucked up to ~30 GB of RAM.

Note that neither the mailing lists, nor the downloads are affected, as both are on other machines.

It's 9:30pm here now, so I just interrupted the cache purging and let the maintenance script finish without that, so we should be back to normal soon.

Sorry for the inconveniences,