Date: Tue, 20 Nov 2012 11:14:27 +1100 From: Lawrence Stewart <lstewart@freebsd.org> To: Alexander Motin <mav@freebsd.org> Cc: freebsd-fs@freebsd.org Subject: Re: graid often resyncs raid1 array after clean reboot/shutdown Message-ID: <50AACB63.6090608@freebsd.org> In-Reply-To: <508F1045.60002@freebsd.org> References: <508E0C3F.8080602@freebsd.org> <508E3E81.9010209@FreeBSD.org> <508E49AD.4090501@FreeBSD.org> <508E91CF.5070003@FreeBSD.org> <508F1045.60002@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 10/30/12 10:24, Lawrence Stewart wrote: > Hi Alexander, > > On 10/30/12 01:25, Alexander Motin wrote: >> On 29.10.2012 11:17, Alexander Motin wrote: >>> On 29.10.2012 10:29, Alexander Motin wrote: >>>> Hi. >>>> >>>> On 29.10.2012 06:55, Lawrence Stewart wrote: >>>>> I have a fairly new HP Compaq 8200 Elite desktop PC with 2 x 1TB >>>>> Seagate >>>>> ST1000DM003 HDDs in raid1 using the on-board Intel Matrix RAID >>>>> controller. The system is configured to boot from ZFS off the raid1 >>>>> array, and I use it as a KDE GUI (with on-cpu GPU + KMS) desktop. >>>>> >>>>> Everything works great, except that after a "shutdown -r now" of the >>>>> system, graid almost always (I believe I've noted a few times where >>>>> everything comes up fine) detects one of the disks in the array as >>>>> stale >>>>> and does a full resync of the array over the course of a few hours. >>>>> Here's an example of what I see when starting up: >>>> >>>> From log messages it indeed looks like result of unclean shutdown. I've >>>> never seen such problem with UFS, but I never tested graid with ZFS. I >>>> guess there may be some difference in shutdown process that makes RAID >>>> metadata to have dirty flag on reboot. I'll try to reproduce it now. >>> >>> I confirm the problem. Seems it happens only when using ZFS as root file >>> system. Probably ZFS issues some last moment write that makes volume >>> dirty. I will trace it more. >> >> I've found problem in the fact that ZFS seems doesn't close devices on >> shutdown. That doesn't allow graid to shutdown gracefully. r242314 in >> HEAD fixes that by more aggressively marking volumes clean on shutdown. > > Thanks for the quick detective work and fix. I'll merge r242314 back to > my local stable/9 tree and test it. I've rebooted the machine a few times now and the array has been started in optimal state without requiring a rebuild each time. Thanks again for the fix. Cheers, Lawrence
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?50AACB63.6090608>