|
From: Folkert Rienstra on 22 Jan 2008 13:19 Arno Wagner wrote in news:5vjpclF1mv66iU1(a)mid.individual.net > Previously Robert Wegner <robweg(a)gmx.net> wrote: > > Hi there, > > > i've got an old storage server here, running W2k, with a promise > > pci-card raid controller. The controller hosts a raid 5 array and it > > is 4 or 5 years old, so i guess it isnt built any more. So, i wonder, > > what if the controller fails and i can't buy exactly the same one? > > Can i simply build in any new raid5-capable controller-card? > > Or does the controller has to be compatible in some way (to the > > old card)? > > The problem is that the RAID superblock is not standardized and > hence it it unlikely to work with a different controller, unless the > manufacturer specifically claims compatibility with your present > controller. In fact not even the place the RAID superblock is found > in, is standardized. That is also the reason why, when using hardware > RAID, you should have a spare controller of the same model. > > > Or is the data lost? > RAID5 is easy to recover for professionals. Any RAID is, if you know how to interprete the Raid Controller metadata. > There is also an other > option: dmraid (part of Linux software RAID) can assemble and use > arrays from several different fake-RAID (software-RAID on a card, > which is likely what you have) controllers. Like there is a huge difference on the disk side between hardware and firmware assisted RAID but not between the firmware assisted RAIDs. > This may allow imaging the RAID under Linux, regardless of filesystem > on it. > > My advice is to replace the controller after a full backup and to recre- > ate with a new controller, of which you get two in order to have redund- > ancy. Or to move to software-RAID, which does not have this limitation. > > Arno
From: Arno Wagner on 29 Jan 2008 07:14 Previously DonLogan <navajo(a)neonfeather.com> wrote: > Arno Wagner <me(a)privacy.net> wrote: >>Previously Robert Wegner <robweg(a)gmx.net> wrote: >>> Hi there, >> >>> i've got an old storage server here, running W2k, with a promise >>> pci-card raid controller. The controller hosts a raid 5 array and it >>> is 4 or 5 years old, so i guess it isnt built any more. So, i >>> wonder, what if the controller fails and i can't buy exactly the >>> same one? Can i simply build in any new raid5-capable >>> controller-card? Or does the controller has to be compatible in some >>> way (to the old card)? >> >>The problem is that the RAID superblock is not standardized and hence >>it it unlikely to work with a different controller, unless the >>manufacturer specifically claims compatibility with your present >>controller. In fact not even the place the RAID superblock is found >>in, is standardized. That is also the reason why, when using hardware >>RAID, you should have a spare controller of the same model. >> >>> Or is the data lost? >> >>RAID5 is easy to recover for professionals. There is also an other >>option: dmraid (part of Linux software RAID) can assemble and use >>arrays from several different fake-RAID (software-RAID on a card, >>which is likely what you have) controllers. This may allow imaging the >>RAID under Linux, regardless of filesystem on it. >> >>My advice is to replace the controller after a full backup and >>to recreate with a new controller, of which you get two in >>order to have redundancy. > Yes he has a working Raid & wants to migrate before failure. And he's > waited too long. > So how does he do this backup / replace controller / use two? > "recreate"? Not easy. Or do Promise offer an data upgrade or > compatibility option? Very simple: For Backup he uses his standard backup method. If he does not have one, the data is obvisouly worthless, and he can just erase it. (Well, or right now establish a backup nethod....) >> Or to move to software-RAID, which does >>not have this limitation. > again - migration path? See above. Remember, RAID is abaout availability. It is not substitute for backup. Arno
First
|
Prev
|
Pages: 1 2 3 Prev: Data missing from MyBook Next: Maxtor ext hd error--corrupted or unreadable |