Re: inconsistent state after crash recovery

On Fri, Jul 26, 2013 at 8:27 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Andres Freund <andres(at)2ndquadrant(dot)com> writes:
>> On 2013-07-26 13:33:13 +0900, Satoshi Nagayasu wrote:
>>> Is this expected or acceptable?
>
>> I'd say it's both.
>
> Postgres is built on the assumption that the underlying filesystem is
> reliable, ie, once you've successfully fsync'd some data that data won't
> disappear. If the filesystem fails to honor that contract, it's a
> filesystem bug not a Postgres bug. Nor is it reasonable to expect
> Postgres to be able to detect every such violation. As an example,
> would you expect crash recovery to notice the disappearance of a file
> that was touched nowhere in the replayed actions?

Eh, maybe not. But should we try harder to detect the unexpected
disappearance of one that is?

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company