On Thu 22 Oct 2009 16:00:39 NZDT +1300, Roger Searle wrote: > confirmation of being on the right track to resolve this. Seems from the > failure of more than 1 partition I will be needing to buy a new disk
Absolutely. If the number of badbocks suddenly goes up, it's a paperweight. You should get a warranty replacement if there are any bad blocks during the wty period. Support your claim with evidence, aka smartctl -a. > The failed disk is 320GB, and contain (mirrored) /, home, and swap. > Presumably I could buy much larger disks, and need to repartition prior to > adding it back into the array? > The partitions should be at least the same size but could be much larger > without any problem? If you were to replace a failed disk in a raid1, remove the failed disk physically, install the new disk physically, boot, and partition the new disk the same way as the one you removed was partitioned. You do NOT have to raid the whole disk. You DO HAVE to make the raided partitions of equal size. No they do not have to be at exactly the same location at each disk with Linux kernel raid, but there are advantages in doing so when it comes to boot loader installation. When you have created partitions matching the existing raid partitions in size, use mdadm to add them into the raid. They'll be synced automatically. See /proc/mdstat I've done this a couple of times with zero trouble when a disk blew up. I'll never have a Linux desktop without 2 disks in raid1 again. You should not need to run mdadm for anything else. In your case, as Steve pointed out, if both your raid1 disks are from the same batch, they'll both fail at the same time. One's already history. You're living on borrowed time. Mind you, I had fun once testing out how long a disk would keep on going which by all means was dead ten times. Years, if I remember. No I did not have anything important on it. Just don't bother complaining when the smoke goes up. Conclusion: when making a raid, never buy all disks from the same manufacturer. (This rule may not be valid for high performance SCSI server disks; it definitely is for consumer junk models. Note disks come in two flavours: high performance server, and consumer junk. You easily tell them apart by price.) You can not enlarge existing raids. You can buy bigger disks and add another raid partition set. As Steve says, you probably want to just buy two bigger disks (no, not from the same make), stick them in, copy the old onto the new, and pull the old. Oh, and set up smartd. Only idiots don't, IMHO. No guarantees it'll save your bacon, but much better than nothing. If you want to improve matters, lod smartctl -a twice daily and write a script to tell you if the bad blocks, pending, or uncorrectable numbers go up. Volker -- Volker Kuhlmann is list0570 with the domain in header http://volker.dnsalias.net/ Please do not CC list postings to me.
