From: nospam@needed.invalid   
      
   On Sat, 2/14/2026 4:29 AM, Mike Scott wrote:   
   > On 13/02/2026 22:03, Lawrence D’Oliveiro wrote:   
   >> On Fri, 13 Feb 2026 13:34:42 -0500, Alan K. wrote:   
   >>   
   >>> Acronis can do incremental, but I don't really need it for that.   
   >>   
   >> The nice thing about rsync is, its --link-dest option lets you create   
   >> incremental backups that look like full backups for restoration   
   >> purposes. It saves space by not making additional copies of files that   
   >> haven’t changed, but because standard POSIX/*nix filesystems allow the   
   >> same file to be linked into multiple directories, that same file can   
   >> appear in multiple snapshots taken at different times, and restored   
   >> from any of them.   
   >   
   > That's a mixed blessing. If the single set of blocks in the backup gets   
   corrupted, all "copies" are affected. Unlike a normal backup.   
   >   
      
      
   Some products support Verify, which allows you to detect some failure   
   situations.   
   You can repeat the backup sequence to two drives, if you consider this   
   failure condition to be a factor.   
      
   On one machine, I was randomly running Verify on some things, when I got a   
   failure.   
   I checked backwards in time, and perhaps two other backups also failed.   
      
   And this was traceable to bad RAM. Once the RAM was tested (and the whole   
   set of DIMMs replaced), the Verify errors stopped happening. The RAM error   
   seemed   
   to be in low memory, but I could not locate it by testing each DIMM   
   individually   
   (with memtest). Bus loading effects can account for this (contributing factor).   
      
   In general terms, a Full+Incrementals tends to have a higher failure   
   rate anyway, than a Full alone. While there is an implementation   
   of Incrementals-Forever (which "synthesizes" a Full from a   
   previous set of Full+Incremental), it is unclear how thorough   
   a job that does of vetting materials along the way.   
      
   Fulls are expensive, but there's no question they get the job done.   
   There is less guess-work, and the scheme could even have a Verify   
   for you (restore to /dev/null).   
      
   I don't think I've ever had a Verify failure on a backup, that was   
   traceable to a bad block. My shitty drive collection, those   
   drives were recognizes early on as less than saintly, and they   
   never got used for backing up anything important. I have   
   around eight hard drives, and ten SSDs, that would not be   
   rated as "backup holders", due to the fact they already   
   show signs of being shitty items. The SSDs are TLC based   
   (which is OK), which seem to lack firmware to "freshen"   
   blocks which have become mushy (which is bad). Just because   
   a drive is cheap, does not mean the firmware has to be   
   bad as well. Another manufacturer learned the lesson   
   about TLC maintenance and offered a firmware flash to help.   
      
   And as hard drive prices rise, there's not much chance   
   the people who really need backups, are going to get them now :-)   
   For example, there is "a" backup drive in town, for $589. Bargain.   
      
      Paul   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)