Il 30/09/20 05:10, Olivier ha scritto: > But if you don't use RAID, you can double the number of backups you are > keeping, that includes older versions. That is a risk calculation > between having secure backups and loosing the backups data as same time > as the primary data. We are using a 3U enclosure with 16 hot-swap bays. The SSD for SO is mounted internally and all the hot swap bays are populated w/ 4TB WD RED disks configured in RAID6 via mdadm (I don't trust "HW RAID", since in case of a controller failure your data is usually toast, but you can always find some way to connect a bunch of drives to a mobo). In case of a disk failure the only real problem is identify which is the failed disk (just run "md5sum /dev/mdX" and see which one isn't working, unless you know how to handle enclosure LEDs).
> No, but I never dug that very much. i need to replace a disk or mount > and older disk about twice a year, so I can reboot at that time. I always prefer hot-swap bays. The only thing I should have changed is not to use all the slots: a free slot is useful both for connecting an archive volume for offline storage and to rebuild the array in a safer way (in some cases, the 'bad' disk is still mostly readable, so md can copy data from it instead of having to read all the other disks, which could expose another failing disk). > Developpers can also be told to let their workstation on all the time, > they can understand the reason why, then you can use > Amanda. Administrtive staff do not always understand these details. Administratives can be taught (but usually only the hard way) that the only recoverable documents are the ones they put on the dedicated network share. ]:) -- Diego Zuccato DIFA - Dip. di Fisica e Astronomia Servizi Informatici Alma Mater Studiorum - Università di Bologna V.le Berti-Pichat 6/2 - 40127 Bologna - Italy tel.: +39 051 20 95786
