Re: Write Ahead Logging for Hash Indexes

On Wed, Mar 15, 2017 at 12:53 AM, Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> * Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
>> Stephen Frost <sfrost(at)snowman(dot)net> writes:
>> > * Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
>> >> It's true that as soon as we need another overflow page, that's going to
>> >> get dropped beyond the 2^{N+1}-1 point, and the *apparent* size of the
>> >> index will grow quite a lot. But any modern filesystem should handle
>> >> that without much difficulty by treating the index as a sparse file.
>>
>> > Uh, last I heard we didn't allow or want sparse files in the backend
>> > because then we have to handle a possible out-of-disk-space failure on
>> > every write.
>>
>> For a hash index, this would happen during a bucket split, which would
>> need to be resilient against out-of-disk-space anyway.
>
> We wouldn't attempt to use the area of the file which is not yet
> allocated except when doing a bucket split?
>

That's right.

> If that's the case then
> this does seem to at least be less of an issue, though I hope we put in
> appropriate comments about it.
>

I think we have sufficient comments in code especially on top of
function _hash_alloc_buckets().

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com