Different ways of measuring catalog size?

Since those are DC tables maybe you don't have the same amount data in the dc_allocation_pool_statistics_by_second vs dc_allocation_pool_statistics?

What do see for minimum TIMEs in those tables?

SELECT node_name, MIN(time) FROM dc_allocation_pool_statistics_by_second GROUP BY 1 ORDER BY 1;

SELECT node_name, MIN(time) FROM dc_allocation_pool_statistics GROUP BY 1 ORDER BY 1;

Also, what is the METADATA pool reporting?

Example:

dbadmin=> SELECT current_value, description FROM configuration_parameters WHERE parameter_name = 'EnableMetadataMemoryTracking';
 current_value |                       description
---------------+----------------------------------------------------------
 1             | Track memory used by Vertica metadata in a resource pool
(1 row)


dbadmin=> SELECT node_name, memory_size_kb, memory_size_actual_kb
dbadmin->   FROM resource_pool_status
dbadmin-> WHERE pool_name = 'metadata';
    node_name     | memory_size_kb | memory_size_actual_kb
------------------+----------------+-----------------------
 v_vmart_node0001 |          94516 |                 94516
 v_vmart_node0002 |         191823 |                191823
 v_vmart_node0003 |         120851 |                120851
(3 rows)

This is the query I've always used to calculate catalog size.

\set fmt12 '\'999,999,999,999\''
select c.node_name,
to_char(c.SAL_total, :fmt12) as SAL_total,
to_char(c.SAL_free, :fmt12) as SAL_free,
to_char(c.Global_total, :fmt12) as Global_total,
to_char(c.Global_free, :fmt12) as Global_free,
to_char(c.SAL_total + c.Global_total, :fmt12) as Total,
to_char(c.SAL_free + c.Global_free, :fmt12) as Free,
to_char(c.SAL_total + c.Global_total -
c.SAL_free - c.Global_free, :fmt12) as Used
from (
select a.node_name,
max(decode(a.pool_name, 'Global', free_memory, NULL)) as Global_free,
max(decode(a.pool_name, 'Global', total_memory, NULL)) as Global_total,
max(decode(a.pool_name, 'SAL', free_memory, NULL)) as SAL_free,
max(decode(a.pool_name, 'SAL', total_memory, NULL)) as SAL_total
from (
select b.node_name, b.time, b.pool_name, b.total_memory, b.free_memory
from dc_allocation_pool_statistics b
join (select node_name, pool_name, max(time) as maxtime
from dc_allocation_pool_statistics
group by node_name, pool_name) as maxes
on (b.node_name=maxes.node_name and b.time=maxes.maxtime)
) a
group by a.node_name
) c
order by c.node_name ;

But like yours, it can sometimes return wildly skewed results from different nodes. I think some of the data doesn't always get purged from certain nodes right away, so it can linger for a while in that state.

OK, I'll have them try some of these other queries, but maybe we should update vBuddyLite to check metadata pool and vs_allocator_usage too?

FYI...

SAL = Storage Abstraction Layer

Anyone remember SAL 9000 from 2010: The Year We Make Contact? She was HAL's twin appearing as a blue camera eye instead HAL's red one.