Date: Fri, 22 Feb 2008 12:56:34 -0500 (EST) From: "Brian A. Seklecki" <lavalamp@spiritual-machines.org> To: Wojciech Puchar <wojtek@wojtek.tensor.gdynia.pl> Cc: Gunther Mayer <gunther.mayer@googlemail.com>, questions@freebsd.org, Bill Moran <wmoran@potentialtech.com>, Chad Ziccardi <cz@digitalfreaks.org> Subject: Re: automatic fsck on gmirror failure Message-ID: <20080222124909.Q83174@arbitor.digitalfreaks.org> In-Reply-To: <20080222181417.F4445@wojtek.tensor.gdynia.pl> References: <47A6287C.9040902@gmail.com> <20080203233311.X7800@wojtek.tensor.gdynia.pl> <1203698632.10391.130.camel@soundwave.ws.pitbpa0.priv.collaborativefusion.com> <20080222181417.F4445@wojtek.tensor.gdynia.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 22 Feb 2008, Wojciech Puchar wrote: >> gmirror(8) / geom(8) should automatically remove (degrade) components >> with bad I/O operations after a certain threshold, but I'm pretty sure >> it doesn't. > > but i'm absolutely sure it does because it did several times for me > Finally I had some time to research. 939 of geom_mirror -- kern.geom.mirror.disconnect_on_failure -- It's a newer 6.x thing apparently: Behavior is not tunable. It happens on a single failure. I ask about tunable behavior because some cheap IDE (Maxtor) disks can fail, then recover. 6.3/amd64: [root@violenjiger-vm /usr/src-RELENG_6_3]# sysctl -a|grep -i kern.geom.mirror kern.geom.mirror.sync_requests: 2 kern.geom.mirror.disconnect_on_failure: 1 kern.geom.mirror.idletime: 5 kern.geom.mirror.timeout: 4 kern.geom.mirror.debug: 0 But: FreeBSD wingspan 5.5-RELEASE-p10 FreeBSD 5.5-RELEASE-p10 #0: Fri Jan 12 seklecki@wingspan:/home/seklecki$ sysctl -a|grep -i kern.geom.mirror kern.geom.mirror.debug: 0 kern.geom.mirror.timeout: 0 kern.geom.mirror.idletime: 5 kern.geom.mirror.reqs_per_sync: 5 kern.geom.mirror.syncs_per_sec: 1000
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20080222124909.Q83174>