Re: half-atomically rebuild databases

[Kristaps Dzonsons <kristaps@bsd.lv>]

> Joerg Sonnenberger wrote on Sun, Jan 01, 2012 at 04:55:08PM +0100:
>> On Sun, Jan 01, 2012 at 04:45:38PM +0100, Ingo Schwarze wrote:
>
>>> This is not perfect because the mandocdb process might get -KILLed
>>> between the two renames, but i don't see a syscall anywhere to
>>> atomically rename *two* files.
>
>> Why do you have to?
>
> There are two files to rename(2):  whatis.{db,index}.
>
>> You can create the new file under temporary name in
>> the same directory and rename it to the normal file name.
>
> That's exactly what the patch i sent does:
> Creating both files under temporary names in the same directory
> and rename both to the normal file names when both are ready.

Hi Ingo,

I don't like this: no matter how we splice it, at some point the files 
will be inconsistent.  I also don't think it's unreasonable to have a 
"bogus" database for a few seconds when they're being rebuilt so long as 
this is documented.

However, if this is a problem scenario, we can just use flock(2) and be 
upfront that the database is inconsistent (apropos_db.c and mandocdb.c 
would need to sync open order) instead of crossing our fingers and 
hoping that they don't dbopen(3) between rename(2)s.

Thoughts?

Kristaps
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv