Date: Fri, 21 Jan 2005 07:55:48 -0500 From: Bart Silverstrim <bsilver@chrononomicon.com> To: FreeBSD Questions <freebsd-questions@freebsd.org> Subject: Re: Hardware RAID Message-ID: <C6026F14-6BAB-11D9-AF87-000D9338770A@chrononomicon.com> In-Reply-To: <20050121090216.GC31548@pcwin002.win.tue.nl> References: <41ED86F4.1080301@mykitchentable.net> <LOBBIFDAGNMAMLGJJCKNAEBEFAAA.tedm@toybox.placo.com> <16879.45212.984246.42585@szamoca.krvarr.bc.ca> <20050121090216.GC31548@pcwin002.win.tue.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Jan 21, 2005, at 4:02 AM, Stijn Hoop wrote: > On Thu, Jan 20, 2005 at 05:22:36AM -0800, Sandy Rutherford wrote: >>>>>>> On Wed, 19 Jan 2005 22:57:21 -0800, >>>>>>> "Ted Mittelstaedt" <tedm@toybox.placo.com> said: >> >>> This did teach me a lesson that I kind of knew already but >>> didn't think too much about. That is, a software array is no >>> substitute >>> for a hardware array. ... > > I respectfully disagree here; it is a substitute in some respects, > especially if you factor in cost. > > My vinum volumes allowed me to survive for a long time without backups > (bad idea, don't do that), and for the past years have allowed me to > survive without having to restore my backups. This through about 5 > failing ATA disks and multiple upgrades of the storage space. > > I'd say it was worth it for me, including reliability. > > If you need speed, or have the cash, etc, you can go for hardware > RAID. But even there I've seen and heard horror stories of > incompatible disks, spontaneously lost configurations or even worse, > silent data corruption due to a bad disk. Just to interject such a tale, since we just had to put up with it... This was with a Windows 2000 server on a Dell with a Perc 3/di RAID controller, using four drives in a RAID 5 array. We came in to find that one of the disks had "gone bad" and the server was blinking red. Disk 2 was dead. Not a problem, with the Dells with a Perc card you just call it in, they send a new drive, you remove the bad and insert the new and it should start rebuilding! The wonder of hardware RAID...hot swap rebuilding to minimize downtime. Well...it wouldn't rebuild. Go around with the tech a couple times, and then ran the onboard diagnostics on the RAID controller...disk 2 was brand new, of course, so it was blank. Disk 3 kept showing about five bad blocks on it. Turns out that sometimes disks will have bad blocks that the array controller can't repair (even though in the utilities it would run the repair and not give any indication that the repair didn't work), and it didn't warn about the bad blocks either; those bad blocks will prevent the controller from rebuilding the array. The only solution? Make a full backup, replace the other drive as well, then rebuild the volume from scratch and restore your data. But hey, who needs a weekend anyway? :-) Hardware RAID should keep you running for awhile, but in this case, it was only a stopgap to buy some time. Like I said, this just happened to us, so thought I'd share. -Bart
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?C6026F14-6BAB-11D9-AF87-000D9338770A>