Re: a fast bloat measurement tool (was Re: Measuring relation free space)

On Fri, Apr 3, 2015 at 9:07 PM, Abhijit Menon-Sen <ams@2ndquadrant.com>
wrote:
>
> At 2015-03-31 22:43:49 +0530, ams@2ndQuadrant.com wrote:
> >
> > I'm just posting this WIP patch where I've renamed fastbloat to
> > pgstatbloat as suggested by Tomas, and added in the documentation, and
> > so on.
>
> Here's the revised version that also adds the count of RECENTLY_DEAD and
> INSERT/DELETE_IN_PROGRESS tuples to the call to vac_estimate_reltuples.
>

I think you have missed to address the below point:

>> 4) prefetch
>>
>>    fbstat_heap is using visibility map to skip fully-visible pages,
>>    which is nice, but if we skip too many pages it breaks readahead
>>    similarly to bitmap heap scan. I believe this is another place where
>>    effective_io_concurrency (i.e. prefetch) would be appropriate.
>>

> Good point.  We can even think of using the technique used by Vacuum
> which is skip only when we can skip atleast SKIP_PAGES_THRESHOLD
> pages.


Apart from this, one minor point:
+typedef struct pgstatbloat_output_type
+{
+ uint64 table_len;
+ uint64 tuple_count;
+ uint64 misc_count;


misc_count sounds out of place, may be non_live_count?

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