Re: Grid agent and CPU
- From: Rich J <rjoralist3@xxxxxxxxxxxxxxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Mon, 26 Jun 2017 10:50:34 -0500
On 2017/06/24 04:17, Orlando L wrote:
We had some prod servers with grid/cloud agents installed, but realized that
they were a significant Cpu hog in that they were consuming 20-30% of the
cycles in some AIX LPARs with 4-8 cores consistently. I am talking non stop
for hours, days and weeks. Talking to support did not help us much except
'apply a patch' which did not help.
After lots of internet search, it looks like this seems to be an issue with
several tools and specifically grid agent. Do the listers plan for a certain
amount of CPU during architecture phase for monitoring
Which version of the Agent are you running? The only metric on my
12.1.0.5 Agents on AIX 7.1 that I have that takes any noticeable CPU
time is a custom one I created to inspect a largish V$BH. And I only
collect that every 10 minutes, so the overhead is manageable for me.
This is on a database server with 4 logical CPUs in an LPAR with an
SMT-4 configuration.
Support should be able to help you determine what's causing the CPU
usage. As has already been mentioned, review any metrics that are being
collected every minute. You may also want to check the Agent's logs and
trace files (e.g. gcagent.log), although weeding through the "usual"
Java errors and what may be an out-of-the-ordinary error may take some
work.
GL!
Rich
Other related posts:
