Queries getting a suboptimal plan and then consuming vast amounts of temp due
to poorly sized workareas?
Real time SQL monitoring or raw ASH data will confirm which particular lines of
the bad plan were responsible.
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> on behalf
of Frank Gordon <frankagordon@xxxxxxxxx>
Sent: Thursday, December 6, 2018 10:05:16 AM
Subject: Rare huge surge in TEMP space usage
Oracle 18.104.22.168.0, Windows.
A query that ran properly for months, for a single day surged in it's
usage of TEMP space, we increased the TEMP space 5x and had to wait
for more disk to be allocated. During the waiting period the query
went back to normal.
The query is of the form
WITH t AS (select ... from tables WHERE ... group by everything),
u AS (select ... from T, table WHERE ..)
SELECT ... FROM tables, T, U
The "group by everything" is suspicious; as is the usage of the same
tables over and over again within the subquery factoring clauses.
Materialized views and subquery factoring also suspicious. HASH JOINS
In the plan a CARTESIAN JOIN is suspicious also.
Started looking at historical stats and 10046/10053 traces.
Anyone ever seen anything like this before?