Re: Debian 10 boot problems with corrupted rw /var

[Richard Weinberger <richard.weinberger@xxxxxxxxx>]

Travis,

On Tue, Aug 20, 2019 at 7:17 PM Travis Griggs <travisgriggs@xxxxxxxxx> wrote:
> 1. How can I get to the bottom of what/why my /var file system is getting corrupted?

This sounds a little like the UBIFS xattrs issue I've been hunting
down. Fixes are upstream but not in stable-trees yet.
Can you please give Linus' tree as of today a try?

> 2. I accept that UBIFS doesn't protect me from power-cycle corruption issues 100%. I have been lead to believe it simply makes it more robust in the presence of power cycle events. But why does it always fail between 1600 and 1800 iterations? If it was truly random, I'd expect it to fail at random power cycle events. It's like some counter is wearing out though.

UBIFS should not die from power-cuts. I have no idea why it fails in
your case between 1600 and 1800.

> 3. The assumption I'm working on here is that having the majority of the system be read-only would increase the robustness of the system (less brick able). But it seems it's only made it so it boots further, but ultimately still bricks. What can I add it to make the ro/rw split actually meaningful?

Well, the boot fails because your system depends hard on a rw UBIFS?
Usually such a split is useful to make update concepts easy or to
detect bad programs.

> 4. It’s been suggested that I should place the two separate volumes (rootfs and /var on separate mtds). I’m going to experiment with that, but will it make a difference?

Never have multiple UBI instances on the same flash. UBI should use as
much from the flash as it can to have a large
wear leveling domain. On top of UBI you can have multiple volumes if you want.

-- 
Thanks,
//richard

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/