seq scan in the case of max() on the primary key column

Hi everybody,

I am running PostgreSQL 9.0 which performs well in most of the cases. I would skip all the parameters if these are not necessary.

I need to frequently (every min) get the max value of the primary key column on some tables, like this case which works perfectly well:

But I have the following similar case which surprises me quite a lot:

explain analyze select max(createdtime) from appqosdata.tcpsessiondata;

-------------------------------------------------------------------------------------------------------------------------------------------------------------------
Aggregate (cost=1123868.30..1123868.31 rows=1 width=8) (actual time=376932.636..376932.637 rows=1 loops=1)
-> Append (cost=0.00..965113.04 rows=63502104 width=8) (actual time=0.020..304844.944 rows=63501281 loops=1)
-> Seq Scan on tcpsessiondata (cost=0.00..12.80 rows=780 width=8) (actual time=0.002..0.002 rows=0 loops=1)
-> Seq Scan on tcpsessiondata_default tcpsessiondata (cost=0.00..965100.24 rows=63501324 width=8) (actual time=0.015..173159.505 rows=63501281 loops=1)
Total runtime: 376980.975 ms

I have the following table definitions:

CREATE TABLE appqosdata.tcpsessiondata_default
(
 Primary key(createdtime), --bigint
check (sessionid >= 0),

 Foreign key(detectorid, sessionid) References appqosdata.tcpsessions(detectorid,id)
 
) inherits (appqosdata.tcpsessiondata);

CREATE TABLE appqosdata.tcpsessions
(
detectorid smallint not null default(0) references appqosdata.detectors(id),
id bigint not null,

 ...

primary key(detectorid, id)
);

As you can see I have tens of millions of rows in both tables which would be ten times more in production. So seq scan is not acceptable at all to get one single value.

Why that difference and what can I do to make the first query use its index on the primary key.

Thank you,

Svetlin Manavski