Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Were you using decent microSDs? Was each mSD getting QC'ed? We benchmark every single mSD before deploying an SBC with it, and have switched to only using Evo+ 32GB cards as they pass our QC 3/5 of the time, which is much better than the other available brands.

Also, mSD corruption means your using a mSD card with a bad controller. The mSD was bound to corrupt and was a low caliber card to start with, unless you are doing a bunch of writes a mSD should last 2 years minimum, if not longer.

Another tact to take would be to switch to an OrangePi PC+ (8GB eMMC included, $22) or similar, that way you know your getting good quality nand with a decent controller out of the box.



Do you mind sharing your quality control method for this? We are hitting card issues on relatively low device numbers.


Just saw this, TL;DR we transfer a large file onto and off of the mSD card. If its above 10MB/s for both (which is the base rating on a class 10 card), then it is a fine mSD card. I've also noticed that on multi-gig files speeds tend to drop to around 6MB/s, not the end of the world for us, but also not a sign of a bad mSD either.


From my experience the failures were almost always caused by a power loss.

The obvious solution would have been to include a battery backup in the case.


Odd, I've only ever seen this be an issue on the Raspberry Pi platform, outside of that power loss doesn't regularly cause data loss on most Androids or Allwinner boards so long as the mSD isn't of low caliber.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: