Brion Vibber wrote: Started it a couple hours ago. There were a couple of issues combining: 1) The main dumps sever (benet) crashed on December 6. It was rebooted, but the monitor process that updates the progress list wasn't restarted. As a result, dumps continued running (in a single worker thread on another box), but without being listed on the status page. 2) One of our secondary file servers (storage2) had filled up; this contains both a backup of the wiki file uploads (from amane) and the dump data for German Wikipedia. We're starting to split upload data between the main server (amane) and a secondary (storage1) due to pressing against the size limits, so I'm also trimming down the on-site backup on storage2. For now I'm removing the thumbs from the backup, as they aren't really necessary (though it'd be nice to not have to regenerate them all in a failover situation!) With space available again, I restarted the German Wikipedia dump on a new worker thread. -- brion vibber (brion @ wikimedia.org)

On Thu, Dec 20, 2007 at 04:12:51PM -0800, Brion Vibber wrote: For what it's worth, that sounds like the same set of symptoms I saw last year on a drive on my sister's MythTV box. One drive, if you accessed anywhere in a specific range of sectors, would knock the IDE controller offline: you could no longer access *anything on any drive*, until you rebooted, at which point you were golden, until you accessed those sectors again. Worked on 2 different motherboards, with 2 different chipsets, under Linux, Windows, and Spinrite's FreeDOS. Clearly it was hardware, but I've never seen it happen ever any other time in 20 years. Cheers, -- jra -- Jay R. Ashworth Baylink jra(a)baylink.com Designer The Things I Think RFC 2100 Ashworth & Associates http://baylink.pitas.com '87 e24 St Petersburg FL USA http://photo.imageinc.us +1 727 647 1274 Witty slogan redacted until AMPTP stop screwing WGA