The following bug has been logged on the website:
Bug reference: 9735
Logged by: Maxim Boguk
Email address: maxim.boguk@gmail.com
PostgreSQL version: 9.3.3
Operating system: Linux
Description:
Hi,
An application start getting curious error during simple query:
select * from transactions where (date(provider_tx_datetime) =
'2014-03-13'::date);
ERROR: tuple offset out of range: 0
Plan:
Bitmap Heap Scan on transactions (cost=614.04..32037.39 rows=167485
width=608)
Recheck Cond: (date(provider_tx_datetime) = '2014-03-13'::date)
-> Bitmap Index Scan on transactions_tx_prv_tx_dt_date_idx21
(cost=0.00..572.17 rows=167485 width=0)
Index Cond: (date(provider_tx_datetime) = '2014-03-13'::date)
What doesn't helped:
set vacuum_freeze_table_age to 0;
vacuum freeze verbose transactions;
and rebuilding the transactions_tx_prv_tx_dt_date_idx21 index.
What helped:
set enable_bitmapscan to 0;
and executing the same query again
with plan:
Index Scan using transactions_tx_prv_tx_dt_date_idx21 on transactions
(cost=0.56..33915.12 rows=167485 width=608) (actual time=0.047..1233.998
rows=195824 loops=1)
Index Cond: (date(provider_tx_datetime) = '2014-03-13'::date)
Total runtime: 1247.688 ms
afterward.
This situation repeated more few times over the last days (with diferen
dateranges).
Any suggestions where and what for I should look next?
It seems somehow related with visibility map and bitmap scan interaction
(only my theory though).
Kind Regards,
Maksym
