RMAN runblock without "run" consumes all memory
From: Herring, David <HerringD_at_DNB.com>
Date: Tue, 28 Jan 2014 14:59:26 -0600
Message-ID: <AD8FE6616C097545A4C9A8B0792909AC2E6A8705AB_at_DNBEXCH01.dnbint.net>
I came across a good one today. We had a problem with a server where an rman process tried to consume all memory, causing all swap to be consumed and then eventually crash. Turns out a DBA made a change to an RMAN backup run block to include an additional command. In the process of this change he accidentally dropped the keyword "run". So instead of
Date: Tue, 28 Jan 2014 14:59:26 -0600
Message-ID: <AD8FE6616C097545A4C9A8B0792909AC2E6A8705AB_at_DNBEXCH01.dnbint.net>
I came across a good one today. We had a problem with a server where an rman process tried to consume all memory, causing all swap to be consumed and then eventually crash. Turns out a DBA made a change to an RMAN backup run block to include an additional command. In the process of this change he accidentally dropped the keyword "run". So instead of
"run
{
...
}
exit"
the keyword "run" was removed.
What happens with this is the RMAN commands within the runblock execute but then the process endlessly fetches more memory, apparently looking for what it should do next.
I've confirmed this on 10.2.0.3 (RHEL 4.6) and 11.2.0.3 (RHEL 5.8), catalog and nocatalog.
Dave Herring
-- http://www.freelists.org/webpage/oracle-lReceived on Tue Jan 28 2014 - 21:59:26 CET