Re: Write Ahead Logging for Hash Indexes

On Tue, Aug 23, 2016 at 8:54 AM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> $SUBJECT will make hash indexes reliable and usable on standby.
> AFAIU, currently hash indexes are not recommended to be used in
> production mainly because they are not crash-safe and with this patch,
> I hope we can address that limitation and recommend them for use in
> production.
>
> This patch is built on my earlier patch [1] of making hash indexes
> concurrent. The main reason for doing so is that the earlier patch
> allows to complete the split operation and used light-weight locking
> due to which operations can be logged at granular level.
>
> WAL for different operations:
>
> This has been explained in README as well, but I am again writing it
> here for the ease of people.
>
>
..
> One of the challenge in writing this patch was that the current code
> was not written with a mindset that we need to write WAL for different
> operations. Typical example is _hash_addovflpage() where pages are
> modified across different function calls and all modifications needs
> to be done atomically, so I have to refactor some code so that the
> operations can be logged sensibly.
>

This patch has not done handling for OldSnapshot. Previously, we
haven't done TestForOldSnapshot() checks in hash index as they were
not logged, but now with this patch, it makes sense to perform such
checks.

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