On 6/29/25 3:22 PM, Ethan Azariah wrote:
> On Sat, Jun 28, 2025, at 4:46 PM, Crystal Kolipe wrote:
>> However there is, (separately), a limit to the number of links you can
>> have to the same inode.
>>
>> So if you're using all this to create a kind of versioning filesystem
>> then at least in theory it might eventually hit that limit and stop
>> working, (unless you take steps to prevent that).
> thanks. accounting for that limit would complexify my plans
> excessively. i'd better use another strategy with a different
> filesystem.
>
> i'm thinking i'll use zfs in omnios if my hardware will support it.
> otherwise, either freebsd ufs with its snapshots or plan 9 cwfs because
> i'm familiar with it. reliability issues will be dealt with by my
> planned backup scheme, and initial testing indicates plan 9's O(n)
> directory read performance will be adequate for my usage. at worst,
> it'll be much faster than the ntfs under win10 in too-little ram;
> that's insanely slow.
>
FWIW, nlink_t is uint32_t which gives you quite a bit of time
before it runs out.
In practice, once a thousand snapshots pile up it's far,
far beyond time to remove & store the current filesystem,
image it onto fresh hardware, and start a new cycle.
geoff steckel
No comments:
Post a Comment