That explains it.
Jonathan Lewis schrieb am 06.12.2017 um 09:16:
It's a known feature - at least for parallel queries - that a process that is
inactive for more than 30 minutes is assumed crashed.
MoS Doc ID: 1900001 applies - no workaround suggested, and no obvious
parameters to try tweaking.
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> on behalf
of Thomas Kellerer <thomas.kellerer@xxxxxxxxxx>
Sent: 06 December 2017 08:00:37
Subject: Oracle aborts monitoring a long running statement - how to find the
we are trying to troubleshoot a complicated query that used to run in under
an hour and started to get extremely slow for unknown reasons ("extremely
slow" means, it still isn't finished after 8 hours)
Oracle did not monitor the statement even though it was running very long and
was using parallel execution.
In MOS we found that only statements with a plan shorter then 300 lines are
We were able to change the parameter "_sqlmon_max_planlines" to 500 as
suggested in DocID: 1613163.1.
Now Oracle indeed _starts_ to monitor the statement.
But after an hour or so, the status of the session(s) turns to "DONE (ERROR)"
in v$sql_monitor, but the query keeps on running, so I am confused what the
"ERROR" means there.
My idea was to use the information provided by the real time monitoring to
find the bottleneck of the statement but if it stops, this isn't really
Is there any view where I can find the _reason_ for the "(ERROR)" or the
reason why Oracle stops monitoring the statement?
v$sql_monitor.error_message and v$sql_monitor.error_number are null for the
sessions of that statement
This is on a 126.96.36.199 EE running on Linux
(please don't comment on the old version - this is a hosted environment of
one of our customers. At least an an upgrade to 188.8.131.52 is planned).