Hi Given the features and maturity of md raid, you should use md raid for a production server which uses Linux only. When you want hardware RAID but you cannot afford hardware raid card, get Promise FastTrak. It will serve as IDE port extension card and also give you the best fake hardware raid (ataraid) you can get for a cheap price. FastTrak BIOS tells you which drive failed. RAID 1 (mirrorring) gives you better read speeds and the most up-to-date backup copy possible when one disk crashes. Sometimes to have the most recent snapshot of work is important for certain applications which is why some admins even mirror Swap partitions. When one of the disks crash in an ataraid mirror, you can pass root=/dev/hdeX at the boot time to restore the system and to recover your data. But the best thing is to use FastTrak BIOS to sync data after replacing the bad disk Server redundancy (high availability) is recommended if you have to use ataraid mirror for critical applications (Rich people: Resonate, CSS, Local Director; Poor people: lbnamed, DNS round robin) Recommendations for better mirror diks: Dont use slave disks, Good cooling, Serial ATA if possible, Enable multimode by default (in kernel), Use DMA when available, Use correct idebus=XX at boot time (I dont know who asked but here's the answer) idebus=XX where XX is the PCI clock speed which is 66 MHz for all computers nowadays, it has nothing to do with ATA133 etc. On 26 Jan 2003, Johann Uhrmann wrote: > On Sat, 2003-01-25 at 09:02, Murty Rompalli wrote: > > > > Yes, ataraid and promise proprietary raid both do not support that > > failsafe feature. Hopefully RedHat people will someday integrate such > > features of md raid into ataraid. > > Thanks for that information. I am a little bit confused about > mirroring-ataraid. To make a long story short: > > Why on earth should people use it, if not for that failsafe feature > which is obviously missing? > > > > According to the information I have got so far, ataraid behaves like > this: > > - There are two disks mirrored by a proprietary software driver. > (unless You use the new and experimental open source driver) > The controller itself is just a stupid IDE extension card with > a slightly enhanced BIOS. > > - If one disk fails, the whole system crashes. > > - In that case, the user has to guess which disk failed and copy > it to a spare and restart the machine while praying loud in front of a > wailing wall to the lord of data consistency. > > - The chance of a failure is twice as high as with a single IDE disk. > > - Last but not least: The Promise support is not even worth the time > for writing the support request. > (Maybe that is because they want to keep up the vision of selling > raid hardware.) > > > Could someone disprove those statements, please? > > > Greetings, > > Hans > > > > > > On 22 Jan 2003, Johann Uhrmann wrote: > > > > > On Tue, 2003-01-21 at 19:47, Marcin wrote: > > > > > > > I know I _should_, but in case of failure of one drive, ataraid become > > > > unusable (kernel oopsses). [...] > > > > > > Could someone please verify that kernel oopses occur when one drive on a > > > Promise FT-133 fails? > > > If that was true, then the chance of a system failure would be doubled > > > by the raid instead of being reduced, right? > > > > > > Therefore, a single IDE disk would be more failsafe than a promise raid. > > > I can hardly believe that (well, on the other hand - it's Promise...). > > > > > > > > _______________________________________________ > > Ataraid-list@xxxxxxxxxx > https://listman.redhat.com/mailman/listinfo/ataraid-list >
