On Fri, Jan 26, 2007 at 05:07:28PM +0200, Kristaps Armanis wrote: > Hello info-cyrus, > > Thursday, January 25, 2007, 6:38:17 PM, Jus rakstijat: > >> New setup is 2 boxes each with 2 partitions each ~2T. > RG> What filesystem? > ReiserFS, till know it worked good and fast with hell a lot > of small files. > > RG> With just about any filesystem, I would recommend smaller > RG> partitions (250G or so). It really isn't any harder to > RG> manage, and backup/restore/fsck will be much happier. > > Even if that 2T partition is one 3ware raid controler with > raid5 over some 6 disks? Do you have any idea how long it takes to reiserfsck 2T full of one-file-per-email Cyrus storage? If you guess about a week, you won't be far wrong. This being from having been in the sorry position of having to do so while the users with data on that partition were without service. Never again. 250G on the other hand will take about 6 hours, which is a much more manageable unit of time. This is assuming your problem is filesystem corruption rather than RAID level corruption. We've had more filesystem level problems than hardware level problems though. YMMV. Also, consider data transfer rate vs total storage. This is why we've got a layout where we can restore the data from a failed drive unit to multiple different machines concurrently to reduce the individual downtime if we lose a RAID unit that big. You don't want to be spending days just copying data around before you're restored again. Not if your users are anything like any other email users I've ever seen. Bron. ---- Cyrus Home Page: http://cyrusimap.web.cmu.edu/ Cyrus Wiki/FAQ: http://cyrusimap.web.cmu.edu/twiki List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
