> On Feb 11, 2019, at 11:07 PM, Rowan Worth wrote:
>
> if process A is updating the DB via write() calls, the locking protocol
> ensures that no other process will call write() or fsync(), and in fact not
> even read() (because the DB might be in a partially-updated inconsistent
> state)
On 2/10/19, Edwin Török wrote:
> I was wondering what changes SQLite3 would need in light of the fsync
> problems discovered by the PostgreSQL community
We've been looking into this, even before your message arrived.
When fsync() returns non-zero, that is basically the operating system
telling
On Tue, 12 Feb 2019 at 15:07, Rowan Worth wrote:
> Huh, fascinating stuff. I'm not an sqlite developer but I can shed light
> on some of your questions.
>
> On Tue, 12 Feb 2019 at 09:54, Edwin Török wrote:
>
>> A very conservative interpretation of various fsync bugs in various OS
>> kernels
Huh, fascinating stuff. I'm not an sqlite developer but I can shed light on
some of your questions.
On Tue, 12 Feb 2019 at 09:54, Edwin Török wrote:
> A very conservative interpretation of various fsync bugs in various OS
> kernels [2][5] would suggest that:
>
> #1. the list of known OS issues
Hi,
I was wondering what changes SQLite3 would need in light of the fsync
problems discovered by the PostgreSQL community (see "How is it
possible that PostgreSQL used fsync incorrectly for 20 years, and what
we'll do about it" talk [1]).
[2] lists that MySQL and MongoDB did some changes, has
5 matches
Mail list logo