[OpenAFS-devel] Re: Breaking callbacks on unlink

[Derrick Brashear]

On Wed, Jan 25, 2012 at 4:40 PM, chas williams - CONTRACTOR
<chas@cmf.nrl.navy.mil> wrote:
> On Wed, 25 Jan 2012 13:43:47 -0600
> Andrew Deason <adeason@sinenomine.net> wrote:
>
>> In general? I think it's a very useful behavior for temporary file
>> storage that doesn't linger. If you make a temporary file and just
>> unlink when you're done, it'll stick around if the process is killed
>> halfway through what it's doing. If you open, and then unlink, it goes
>> away if the proc is killed even if you don't cleanup properly. (The
>> OpenAFS salvager was semirecently made to use this kind of functionality
>> for its temp storage, iirc.)
>
> this isnt often used. =A0mostly because it hides the file from someone
> who doesn't know that the program actually does this. =A0in some cases, i
> would almost consider this rogue behavior because you can just fill up
> the filesystem and in the case of some programs, ignore write errors
> and keep it filled until someone figures out what is going on with
> lsof. =A0draw gun, shoot foot.
>
>> For the case in AFS, it's a little weirder to be sharing such files
>> across machines, but it's possible.
>
> i think the case of dso's is a reasonable argument for consistent (and
> in this case unix/posix-like) behavior of unlinked files across clients.
> oddly enough, if you are using emt you wont see this problem because of
> the way it handled upgrades to programs/shared libs. =A0possibly, just to
> avoid this inconsistent behavior.

"it depends". we used emt and i promise you we had problems anyway, albeit =
not
 the one you're concerned about. it depends *how* you use emt. i came up wi=
th a
means to cache dsos and when I left CMU that's what we were using.

(i build a in collection x against so version 1.1 of library l in
collection y; i upgrade collection y, and library l is now so version
2. so now i have a binary which
won't run. solved by simply scanning every binary using ldd and
caching every dependency, and then loading the collection of dsos at
the lowest priority,
so any *real* dso would always win over the cache)

--=20
Derrick