Re: blocking enq-ss wait events

From: Priit Piipuu <priit.piipuu_at_gmail.com>
Date: Thu, 6 May 2021 17:42:07 +0200
Message-ID: <CAJYY02hTnBoY=0zTwYBNQ-HnzmKj=OP9NXFGb_fDdaJ3SwRguw_at_mail.gmail.com>



On Thu, 6 May 2021 at 05:56, Lok P <loknath.73_at_gmail.com> wrote:

> ... "enq-SS contention" ...
>
> The version is 11.2.0.4 of Exadata.
>

In RAC sort segments are handled per instance basis, and space allocated by one instance isn't visible to others. Instead of straight ORA-01652 returned to the client there will be cross-instance call to see if some allocated but unused temp space can be freed. In case of multiple requesters, one will be waiting on DFS lock handle, and other requesters will be serialized on SS enqueue. You can check from GV$TEMP_EXTENT_POOL if there are any large allocations around. This is a bit of a conjecture how things seem to work in 19c, not sure about 11.2 tho...

--
http://www.freelists.org/webpage/oracle-l
Received on Thu May 06 2021 - 17:42:07 CEST

Original text of this message