XPost: aus.computers   
   From: nospam@needed.invalid   
      
   On Sat, 12/27/2025 6:50 PM, keithr0 wrote:   
   > On 25/12/2025 4:57 am, Lawrence D’Oliveiro wrote:   
   >> On Wed, 24 Dec 2025 21:18:19 +1000, keithr0 wrote:   
   >>   
   >>> On 24/12/2025 3:33 pm, Lawrence D’Oliveiro wrote:   
   >>>   
   >>>> I got a new 12TB drive for my backup machine just a couple months   
   >>>> ago.   
   >>>   
   >>> That's a lot of eggs in one basket ...   
   >>   
   >> That’s why I have a backup machine.   
   >>   
   >>> ... personally I'd prefer a RAID 5 or 6 setup with smaller drives.   
   >>   
   >> RAID is about high availability, not about backup.   
   >   
   > RAID is about failure resilience, if you have a single drive a failure loses   
   everything. Make a RAID group, and the loss of a single drive results in no   
   data loss. Done properly, it also improves performance.   
      
   RAID is about delaying maintenance until after the work day is done.   
      
   It's nothing about long term resilience.   
      
   If you have one fault and you do not meet the MTTR, the second   
   fault might knock you out, or depending on RAID class, the third   
   fault might knock you out.   
      
   *******   
      
   You really really have not thought it through,  if you believe that resilience   
   thing.   
      
   At 2PM in the afternoon at work, the main CAD server (serves copies of CAD   
   software), I can't remember if it was a PERC or what it was.   
      
   The *controller* wrote to all drives at once. It was   
   not a commanded write. It was a firmware issue of some sort.   
   And not a capacity-rollover type flaw.   
      
   It corrupted some area low in the disk storage.   
      
   Causing *all volumes to be wiped out instantly*.   
      
   The restore time for the server, ran past 5PM and the   
   five hundred engineers depending on the server for their   
   software, had long since gone home when the server came back up.   
   Somebody in management, did the calc to see what the   
   "lost time" had cost us. That's what was the first step in   
   figuring out what the response should be to the event.   
      
   And this failure happened *twice*, just on a different server   
   which was not as critical. Because the first incident was not   
   as critical, it wasn't taken as seriously and nobody at that   
   time, had figured what had happened to the server. It was only   
   when a major infrastructure incident was raised, that the event   
   was analyzed to its source.   
      
   This is why RAID is worth *nothing* to you.   
      
   There is a class of "common mode faults" you should consider.   
      
   An example of a common mode fault, is when the PSU +12V shoots   
   up to +15V and burns all disk drives. When I mentioned the possibility   
   of this happening, a poster wrote in and said this had happened   
   to him. Loss of all storage on a PSU fault.   
      
   This is *why we do backups of our RAID array* :-/   
      
   A RAID array is NOT a backup.   
      
   A mirror, is just so much bullshit for the people doing it.   
   Because they haven't thought about the common mode faults.   
      
   I can give an example from a USENET posting about a RAID1 two disk mirror.   
      
   A user is running RAID 1. One of the drives drops. The user thinks   
   this is fine and dandy. Plug in replacement drive, and oh... whats this ?   
   The files on the second drive, stopped updating three months ago.   
   The array had some kind of fault, where the controller was   
   not writing to the second (surviving) drive. Which means, sure,   
   you can do a RAID rebuilt of your mirror, but three months   
   worth of changes are missing. The controller in this case was   
   a SIL3112, so it wasn't an actual quality controller. It's just   
   the deception "you have RAID", that counts.   
      
   DO YOUR BACKUPS. And run a Verify on your backup, to ensure it is a good one.   
   I have stories about that, too :-)   
      
      Paul   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)