Conversation

Oh, this is interesting (and a little scary)

tl;dr don’t use SSDs for long term, offline storage. The data degrades after as little as two years without the drives being powered up

https://www.tomshardware.com/pc-components/storage/unpowered-ssd-endurance-investigation-finds-severe-data-loss-and-performance-issues-reminds-us-of-the-importance-of-refreshing-backups

1
19
2
@WiteWulf Flash needs periodic scrubbing, that should be well known? And yes, scrubbing is hard to do with power off :-). Retention on flash also depends on temperature when data was stored and storage temperature, if you want more fun :-).
1
1
7

@pavel it is not well known to the vast majority of users.

1
0
0
@WiteWulf Well, if you want more scary stuff... cut SSD power during write operation. It is quite hard to get FTL right, and I'm pretty sure some vendors simply don't. Result is... a brick.
1
1
3

@pavel @WiteWulf I suspect this is the cause of most rpi sd card corruption, and it's the result of host computer cheaping out on basically one capacitor that could ensure card retains power for a minimum period of time after host has gone down and new commands are no longer possible.

1
0
1
@dalias @WiteWulf Yes, that's likely a problem. And my guess is that a) it is not clear how big capacitor is required and b) rpi expects well-designed cards and cards expects power-supply that does not cut out randomly :-(.
1
0
1

@pavel @WiteWulf The needed size should be easily computable from the max current SD card is allowed to pull and the max write time in the spec or something like that.

1
0
1
@dalias @WiteWulf Its not that simple. SSDs are not that deterministic, and they may need to perform writes even when host does not request writes, due to garbage collection or scrubbing.
0
0
0