If you’re in a panic because the Internet told you that your shiny new SSD may lose data in “just a few days”
The key here is slowly. That article's main focus is talking about the effects of temperature on data integrity in SSDs. Unplugged SSDs do lose data because of entropy. Unless the SSD in question has a battery, it will start to slowly lose data over the course of years.
I didn't panic. I don't even own an SSD. I just linked the debunk article because it tells how some people misinterpreted a study, saying that they leak data, when in fact the study says that an SSD only leaks data after it's reached it's end of life, meaning after the cells have reached their maximum write cycles. When that happens you need to replace it anyway, even if didn't leak data.
That quote's from the article.
Sorry for the delay in responding to your request, life happened.
Here are some preliminary results I have been able to throw together.
All this file transfer was over my gigabit network so that is a huge bottleneck, and I don't think I have anything faster I can connect to my server to truly give a max speed.
27 gb
from server 90MBPS over gigabit
5min 8sec
27gb
to server 80MBPS over gigabit
5min 42sec
Here are some screen shots.
I know this thread is ancient, and I hope I am not being rude by bumping it (is that frowned upon here?). I was wondering what would have happened in Wendell’s testing video if instead of a 3 drive raid 5 it was a 4 drive raid 6? As far as the raid controller or md would know, all 4 drives are still working, but of course some files would be corrupt on one of the disks… Now instead of not knowing to trust the parity or the data, it should either be the case that there’s disagreeing data and parity that agrees with one of the datas… or something like that. I sure would appreciate a run-though of what would be happening in that situation. I would think it should recover itself in this situation. I guess I could test this myself, but I am lazy and am hoping someone who knows will save me the trouble.
sounds like a new video.
A few years ago, these errors were caught and corrected during a scrub (“patrol” in LSI vernacular) but not on the fly, interestingly. We will have to retry in 2018 with modern controllers and firmwares to see what happens. Netro generally is frowned upon because usually it doesn’t add anything to the information.
