> We are not stuck on SATA. The whole data directory has ~ 80GB of data > so PATA would work just as well.
I fear you missed the point. This is hardly about SATA vs. PATA, but rather about ATA vs. SCSI. If you do really need reliability, you should seriously consider SCSI RAID. Yes, it definitely is expensive. If your budget constrains you to ATA, you should seriously consider whether you need RAID at all. Even with RAID, you need good, regular backups. Remember that RAID 1 or 5 is not a backup strategy but just a tool to minimize downtime in case of failure and to avoid the loss of data since the last nightly backup. Is it really required that you do not loose any data, not even of one day, in case of the quite rare event of a *sudden* HD death - remember, most HD failures can be avoided if you watch out for symptoms occurring prior to failure? Or is it really required that the disk is back instantly, i.e. do you really lack the time to manually umount /data sed 's/wd1a/wd2a/' < /etc/fstab > /etc/fstab.new mv /etc/fstab /etc/fstab.old && mv /etc/fstab.new /etc/fstab mount /data Well, you might need to block logins and kill some user processes before doing that, but that's the general idea if you have a nightly clone on a second disk. You can even mount the clone readonly and tell users about it, in case they mistype rm arguments. Let me give an example. For my small LAN of about 15 workstations and about 30 or 40 regular users, i do not really absolutely need that kind of reliability - but i thought it might be a good idea to have RAID 5 in order to be nice to users just in case - and SATA is no more that expensive and all... So about 2 years ago, i got four 160 GB Seagate SATA disks and a cheap SATA hardware RAID controller. Of course, no SATA disks are designed or guaranteed for 24/7 operation, but the dealer specifically told me those would probably stand the strain. In fact, they did not fail. But... Quite similar Seagate SCSI disks from the same period had a firmware flaw triggering occasional failure when given certain SCSI commands in certain combinations. Seagate released a firmware fix, asking its clients to flash the drive BIOSes. The SATA drive BIOS is not flashable, and i do not know whether those disks contain similar bugs. Maybe they do - if so, that might trigger the Adaptec SATA controller failures. Of course i know my Adaptec 2410 SA BIOS is flawed in itself, so i'm considering to replace it by better hardware anyway (LSI Megaraid, whatever). But what if the disks themselves contain BIOS flaws? Maybe those will confuse the better controller as well? Who knows... Perhaps i should just use those SATA disks without RAID, i will probably end up on the same level of reliability and with much less headaches. You see the point? I think this whole mess is still the typical difference of SCSI and ATA. SCSI is designed for reliabily, if it is broken, serious vendors actually try and fix it, but it is expensive. ATA is designed to be cheap. You can run a cheap system with a moderate level of reliabilty using ATA, but don't expect much. Yours, Ingo -- Ingo Schwarze <[EMAIL PROTECTED]>
