Re: Question on Fragmentations
От | Peter Koczan |
---|---|
Тема | Re: Question on Fragmentations |
Дата | |
Msg-id | 45CD01BC.7020305@gmail.com обсуждение исходный текст |
Ответ на | Re: Question on Fragmentations (Michael Monnerie <michael.monnerie@it-management.at>) |
Список | pgsql-admin |
Michael Monnerie wrote: > On Freitag, 9. Februar 2007 04:08 Peter Koczan wrote: > >> Case in point, I use xfs as the filesystem running under postgres, >> and after a few days the "major" database clusters showed ~90% >> fragmentation on their respective partitions (which is about a 10 to >> 1 ratio of file fragments to files). After running a defragmenter >> > > Does xfs have such stats, and defragmenter included? It could be a good > idea for me to use that, then. Currently I use reiserfs. > > mfg zmi > xfs comes with It does have built-in, xfs-approved utilities for stats and defragmenting built-in. xfs_db gives stats (for fragmentation use xfs_db -c frag -r /dev/XXX). This works even if the filesystem is mounted and active, but I believe that old stats are cached until said filesystem is remounted or until some stat collection process runs. xfs_fsr is the defragmenter (simply use xfs_fsr /dev/XXX). It's safe to run this on an active filesystem/database partition, because it throws away the fragmented data if files are changed. So, for full defragmentation, you'll either want to run it offline, unmounted, or during idle times. Peter
В списке pgsql-admin по дате отправления: