Date: Sat, 30 Apr 2011 12:25:21 +0200 From: Andre Albsmeier <Andre.Albsmeier@siemens.com> To: freebsd-rc@freebsd.org Cc: Andre.Albsmeier@siemens.com Subject: New knob for ignoring readonly fss in 340.noid and 310.locate? Message-ID: <20110430102521.GA11716@curry.mchp.siemens.de>
next in thread | raw e-mail | index | archive | help
I have noticed that "periodic weekly" started to consume more and more time on some of my machines. The reason is the amount of snapshots that are used at my site, e.g., on one machine I have two file systems: /dev/ccd0 544G 266G 235G 53% /people /dev/ccd1 556G 372G 140G 73% /share which hold these snapshots: /dev/md12 544G 241G 259G 48% /people/.snap/@GMT-2011.04.26-03.15.04 /dev/md13 544G 241G 259G 48% /people/.snap/@GMT-2011.04.27-03.15.04 /dev/md14 544G 236G 264G 47% /people/.snap/@GMT-2011.04.28-03.15.03 /dev/md15 544G 240G 260G 48% /people/.snap/@GMT-2011.04.29-03.15.03 /dev/md16 544G 238G 263G 47% /people/.snap/@GMT-2011.04.30-03.15.04 /dev/md17 544G 241G 259G 48% /people/.snap/@GMT-2011.04.23-03.15.01 /dev/md18 544G 239G 261G 48% /people/.snap/@GMT-2011.04.16-03.15.01 /dev/md22 556G 344G 167G 67% /share/.snap/@GMT-2011.04.26-03.28.32 /dev/md23 556G 344G 167G 67% /share/.snap/@GMT-2011.04.27-03.28.26 /dev/md24 556G 344G 167G 67% /share/.snap/@GMT-2011.04.28-03.28.42 /dev/md25 556G 341G 170G 67% /share/.snap/@GMT-2011.04.29-03.29.12 /dev/md26 556G 343G 168G 67% /share/.snap/@GMT-2011.04.30-03.26.48 /dev/md27 556G 344G 167G 67% /share/.snap/@GMT-2011.04.23-03.27.33 /dev/md28 556G 341G 170G 67% /share/.snap/@GMT-2011.04.16-03.27.39 /dev/md29 556G 337G 175G 66% /share/.snap/@GMT-2011.04.09-03.22.01 For our purpose, the weekly maintainance is not desired on the snapshots. Of course, it might be possible to modify the behaviour of 310.locate 340.noid by tweaking /etc/locate.rc or weekly_noid_dirs which would be rather annoying since this would have to be done on every machine individually due to their different filesystem names (and pretty sure will be forgotten when fss are added or removed). One condition that would match everything (at least in my case) would be to restrict the work of 310.locate and 340.noid on filesystems which are not mounted readonly. So what would people think about a knob that restricts at least these two scripts (maybe there are more candidates) from letting find(1) decend into rdonly filesystems? Of course, this knob would default to "off" and could be set in periodic.conf and/or locate.rc individually... Thanks, -Andre
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110430102521.GA11716>