On 10/29/10 7:31 AM, Peter Kjellström wrote: > On Friday 29 October 2010 11:42:38 przemolicc at poczta.fm wrote: >> Hi, >> >> we have CentOS ftp server (vsftpd) which has a lot of users who are writing >> and reading a lot of small files from/into its own accounts (and other >> servers - using samba client - are reading these files and putting them >> into outside database). >> Since this server is under heavy load its availability is important. >> >> > From time to time we "crash" this server (don't ask why ...) but then fsck >>> is running for over 20-30 minuts. >> >> The question is: is there any other _stable_ filesystem (xfs ?, jfs ?) >> which we can use instead of ext3 which is (quite) immune to crashes and >> whose fsck is "faster" (by design) then in ext3 ? > > The idea with ext3/ext4 is that you don't have to run a full fsck after a > system crash (only a fully automated journal replay). > > XFS uses the same idea (no fsck only journal replay). But if you really want > to fsck an xfs filesystem then that too will take a lot of time. The question is, are the fsck's happening because the journal is corrupted, because something is wrong with it, or because a journal isn't configured or the 'time to check' has expired. In the latter case you can adjust with tune2fs. -- Les Mikesell lesmikesell at gmail.com