Re: Proposal "VACUUM SCHEMA"

On 12/21/2014 10:30 PM, Fabrízio de Royes Mello wrote:
> [snip]

I do agree that "vacuum schema" might very well be useful (I'll probably
use it myself from time to time, too).
ANALYZE SCHEMA (specially coupled with some transaction-wide "SET
statistics_target" could be beneficial)

>
> > And why that, but not
> > say schema-wide ANALYZE, CLUSTER, TRUNCATE, ...
> >
>
> +1. I can write patches for each of this maintenance statement too.

Hmm... I think Tom might have been a bit rethorical (or even sarcastic
with that), but I can definitely be wrong.

Do we really want to have some such operation potentially (and
inadvertently) locking for *hours* at a time?

CLUSTER SCHEMA somename;

... where schema "somename" contains "myHugeTable"

Given that the cluster command exclusively locks and rewrites the
table, it might lock queries and overwhelm the I/O subsystem for quite a
long time.

TRUNCATE SCHEMA whatever sounds quite dangerous, too.

Just my .02€

/ J.L.