Date: Fri, 23 Nov 2007 15:54:24 +0100 From: "Jan Catrysse" <j.catrysse@proximedia.be> To: <freebsd-questions@freebsd.org> Subject: RE: RAID1 synchronisation - howto OR not necessary? Message-ID: <20071123145054.E383228782@smtp.proximedia.com> In-Reply-To: <20071123092733.3e47283b.wmoran@potentialtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> From: owner-freebsd-questions@freebsd.org=20 > [mailto:owner-freebsd-questions@freebsd.org] On Behalf Of Bill Moran > Sent: Friday, November 23, 2007 3:28 PM > To: Jan Catrysse > Cc: freebsd-questions@freebsd.org > Subject: Re: RAID1 synchronisation - howto OR not necessary? >=20 > "Jan Catrysse" <j.catrysse@proximedia.be> wrote: > > Hi Bill, > >=20 > > Thank you for your input. > >=20 > > I assumed this was common knowledge, but I can be wrong? > > I've checked some other RAID controllers in the company and all of=20 > > them have the need to be verified/synchronized once and a while. >=20 > No. >=20 > There is a _world_ of difference between "verify" and "synchronize". > Periodically verifying the health of a RAID array is good practice. > Re-synchronizing it periodically is stupid. If you have to=20 > do that, then you wasted money on a RAID card. >=20 > > This > > happens in the BIOS for the more expensive cards (> 600=E2=82=AC)=20 > and with a=20 > > utility/driver for the low budget cards... >=20 > Depends on what you're talking about. >=20 > Yes, expensive cards do both health checking and resyncing in=20 > the BIOS without the need of operator intervention. >=20 > Low-end hot-swappable cards will automatically do the=20 > resynchronizing if they detect a HDD change, but often don't=20 > do periodic health checking. >=20 > Low-end cards do neither. However, you have to power the=20 > machine down to replace a failed drive, so you're also=20 > accepting the burden of waiting for the BIOS to resync. It's=20 > part of the cost trade-off. >=20 > > This is what I found in a 3Ware manual: > > Verification can provide early warning of a disk drive=20 > problem or failure. > > ...verification once every 24 hours... > > Not verifying the unit periodically can lead to an unstable=20 > array unit and may cause data loss. > > It is strongly recommended that you schedule a verify at=20 > least 1 time per week. >=20 > Nice documentation ... Do you verify every 24 hours or once a week? >=20 > In any event, the availability of such a utility for FreeBSD=20 > depends on the driver and the (possible) availability of=20 > third-party (or even > vendor-supplied) utilities. For example, LSI provides the=20 > megaraid utility. It's designed for Linux but works on=20 > FreeBSD and allows total control over the RAID card,=20 > including verifications. Reading the man page for the driver=20 > being used may turn up something. >=20 > I don't know if one exists for your specific card, but keep=20 > in mind that the driver you're using may also work with=20 > high-end RAID systems that don't need it, so the absence of=20 > one is possible. >=20 > I suggest you reformat/repost your question with a subject=20 > line more along the lines of "Looking for a control utility=20 > for ICH8R RAID" It's quite possible that the people who know=20 > a lot about that hardware missed the original conversation thread. >=20 > If there is no such utility, I suggest looking into something=20 > like samhain, which will continually validate that the files=20 > on your system are uncorrupted. This has the added advantage=20 > of warning you if someone has cracked your system and=20 > installed a trojan. >=20 > Also consider the benefit of spending the extra $$ on a=20 > high-end RAID card. There are very good reasons that people=20 > are willing to pay more for them. Personally, I wouldn't use=20 > a low-end RAID card ... GEOM would be just as good if not=20 > better, IMHO. Hello again, I understand what you mean by synchronization not beeing the same as = data verification. What I mean is indeed data verification. For the Intel ICH8R controller FreeBSD uses ATA(4), no vendor support is = available. It is possible however to control the array using ATACONTROL STATUS / = DETACH / ADDSPARE / REBUILD / ... but no VERY command seem to exist. As you suggest I will repost the question with another subject. Would it = be a big problem if I contact the FreeBSD developper of ATA(4) directly? I didn't dig in GEOM because I wondered what happens if the primary disk = fails when two disks are in a RAID1 config? Cheers, Jan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071123145054.E383228782>