Обсуждение: PostgreSQL and FreeBSD SoftUpdates
Hi guys, Does anyone else have this problem? We have softupdates turned on on our data dir. (Soon to be turned off due to these issues). The partition is 12GB. 'df' says that we're using 12 and a bit GB but 'du' says we're using 2GB (which we really are). It seems that perhaps softupdates is caching some stuff, or preventing something from being written properly, etc. The funny thing is that this was never a problem until we upgraded to 7.4. Has something changed in the way file writes or syncs are done? Chris
what version of FreeBSD are you using? I'm running 4.9-STABLE with softupdates on my db file system ... On Tue, 30 Mar 2004, Christopher Kings-Lynne wrote: > Hi guys, > > Does anyone else have this problem? > > We have softupdates turned on on our data dir. (Soon to be turned off > due to these issues). > > The partition is 12GB. 'df' says that we're using 12 and a bit GB but > 'du' says we're using 2GB (which we really are). > > It seems that perhaps softupdates is caching some stuff, or preventing > something from being written properly, etc. > > The funny thing is that this was never a problem until we upgraded to > 7.4. Has something changed in the way file writes or syncs are done? > > Chris > > > ---------------------------(end of broadcast)--------------------------- > TIP 8: explain analyze is your friend > ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
> what version of FreeBSD are you using? I'm running 4.9-STABLE with > softupdates on my db file system ... FreeBSD goddard.calorieking.com 4.9-STABLE FreeBSD 4.9-STABLE #2: Mon Jan 26 23:23:17 EST 2004 wg@goddard.calorieking.com:/usr/obj/usr/src/sys/GODDARD i386 We're not 100% sure it's softupdates, but we can't see anything else that it could be. Chris
On Tue, 30 Mar 2004, Christopher Kings-Lynne wrote: > > what version of FreeBSD are you using? I'm running 4.9-STABLE with > > softupdates on my db file system ... > > FreeBSD goddard.calorieking.com 4.9-STABLE FreeBSD 4.9-STABLE #2: Mon > Jan 26 23:23:17 EST 2004 > wg@goddard.calorieking.com:/usr/obj/usr/src/sys/GODDARD i386 > > We're not 100% sure it's softupdates, but we can't see anything else > that it could be. Right off the top of my head, it almost sounds like a file is being held open after its been deleted ... we went through that with the new aspseek a little while back, where 170gig just disappeared overnight, but du showed hardly any disk space being used ... Does restarting the database server (not rebooting, just restarting the postmaster) free up the disk space? ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
> Right off the top of my head, it almost sounds like a file is being held > open after its been deleted ... we went through that with the new aspseek > a little while back, where 170gig just disappeared overnight, but du > showed hardly any disk space being used ... > > Does restarting the database server (not rebooting, just restarting the > postmaster) free up the disk space? No - have to reboot. That's probably because of softupdates though. Chris
On Tue, 30 Mar 2004, Christopher Kings-Lynne wrote: > > Right off the top of my head, it almost sounds like a file is being held > > open after its been deleted ... we went through that with the new aspseek > > a little while back, where 170gig just disappeared overnight, but du > > showed hardly any disk space being used ... > > > > Does restarting the database server (not rebooting, just restarting the > > postmaster) free up the disk space? > > No - have to reboot. That's probably because of softupdates though. 'k, *shouldn't* require a reboot ... but, what I'd try is to do what you've thought .. disable softupdates and see if you can recreate ... if killing off the process auto-reclaims the space fast, then it sounds like a stale file being held open (log file being rotated improperly?) ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
> 'k, *shouldn't* require a reboot ... but, what I'd try is to do what > you've thought .. disable softupdates and see if you can recreate ... if > killing off the process auto-reclaims the space fast, then it sounds like > a stale file being held open (log file being rotated improperly?) ... Log file's on a different partition... Chris
>>> Right off the top of my head, it almost sounds like a file is being >>> held >>> open after its been deleted ... we went through that with the new >>> aspseek >>> a little while back, where 170gig just disappeared overnight, but du >>> showed hardly any disk space being used ... >>> >>> Does restarting the database server (not rebooting, just restarting >>> the >>> postmaster) free up the disk space? >> >> No - have to reboot. That's probably because of softupdates though. > > 'k, *shouldn't* require a reboot ... but, what I'd try is to do what > you've thought .. disable softupdates and see if you can recreate ... > if > killing off the process auto-reclaims the space fast, then it sounds > like > a stale file being held open (log file being rotated improperly?) ... Install the latest version of lsof(8) and see if there are any stale files being held open. I've got databases on FreeBSD 4.X and 5.X with softupdates on both and haven't had a problem. I'm wondering if your database is doing something "exotic" that hasn't been tickled. The first thing that comes to mind is, are you using deferred constraints? Second, if it is a soft updates issue, then a reboot isn't necessary (as Marc says)... you should be able to stop the database and type df -k && sync && sleep 30 && df -k see space being freed up. -sc -- Sean Chittenden