[systemd-devel] Second (erroneous) check of rootfs?

[Nikolai Zhubr]

Hi,
11.01.2015 15:04, Chris Murphy:
> That's all I meant by bouncing back to ext devs. I don't mean there's
> anything wrong with ext4. It's pretty clear the XFS and Btrfs devs
> expect that if a normal rw mount fails, that boot fails and we're
> dropped to a dracut shell with an unmounted root. And hopefully our fs
> repair tool is in the initramfs so we can run it on the unmounted
> root. Is that good enough for ext4 also? Or do they still really want
> e2fsck run every boot?

ext4 is perfectly able to run for years (with moderate everyday load) 
without a single full fsck, I can assure. It is basically a really 
zero-maintenance filesystem (if treated properly).


Thank you,
Nikolai

>
> It just seems there's a more elegant way to do this. Wasn't there some
> big blow up recently, where some people were mad they couldn't cancel
> these lengthy e2fsck's on ext4 volumes? Seems like a bad idea to
> cancel an in-progress fsck anyway, but it had been scheduled 180 days
> prior as the mkfs default. There was already tacit permission to do
> this full fsck. However, a vastly better UX would be to communicate
> the need to do this differently, like a notification that says "a full
> fsck is recommended on the next boot, set this now?"
>
>
>