On Wed, 17 Sep 2003 09:10:10 +0200, Jacob Anawalt wrote: > In other cases, because ext3 is writing it's journal to a disk every few > seconds, it could be a little slower. Since most systems aren't normally > under intense I/O, this is usually unnoticable. You do have the option > of storing the journal to a different disk.
The 2.4 series had some serious problems in that area. I once had several ext3 filesystems mounted and observed regular "load spikes". I.e. the system would jump from 0 to 100% CPU and freeze (!) for about half a second. Then everything goes back to normal again, and a few minutes later comes the next spike. I think it had something to do with the flushing of the ext3 journal -- remounting the same partitions as ext2 fixed it. That behaviour was simply unbearable for a workstation and made me abandon ext3. Maybe the preempt/low latency patches could help nowadays. > I've read people suggesting to not use ext3 over ext2 on some > directories (which means different mounts for those directories.) I > think they talked about places like /var/spool/news but I could be very > wrong. /var/spool/news would probably be better served by ReiserFS or XFS. Those filesystems are a bit faster when it comes to directories which conatain literally thousands of files. I haven't tried both, Reiser because of the ever-repeating stories about data corruption and XFS because it hasn't made its way into the vanilla kernel yet. > Maybe /etc would be better served by a Reiser FS. /etc is too small to make a difference anyway, and is crucial for the system to boot. I'd go the safe route and choose a rock solid ext2 or ext3. -- Best Regards, | Hi! I'm a .signature virus. Copy me into Sebastian | your ~/.signature to help me spread! -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
