This discussion is archived
2 Replies Latest reply: Jan 28, 2013 4:47 PM by 660273 RSS

Sessions with Event "class slave wait" taking 100% CPU

660273 Newbie
Currently Being Moderated
Hi experts,
I have an issue in an 11.2.0.2 RAC database with 2 standby nodes that are driving me nuts.
All started when I've been told there are 2 Oracle processes consuming too much CPU in the environment.

I found 2 processes which takes 100% CPU each. These processes are Oracle processes into the DB and checking them out although they are registered in v$process and v$session, there is nothing related to them in v$bgprocess so I could not find out which oracle processes started these 2 OS processes.

The view v$session shows them ACTIVE in the event "class slave wait". There aren't too much information about this event, at least I failed to find...

Using dbms_monitor and dbms_system to create trace events did not create any tracefile. Only ORADEBUG was able to create events but with nearly no information inside the trace files.

Questions I need to have answered or at least hints that I can follow to find the answer:
1) what is causing the CPUs to be consumed at 100%? Which process?
2) Why does these processes are using that much of CPU?
3) What can be done to safely get over it?

Honestly I don't know where else to look at except trying to get some help. Could someone give me a hand please?

BR,
Lauro Ojeda

Edited by: LauroOjeda on 26/01/2013 06:44
  • 1. Re: Sessions with Event "class slave wait" taking 100% CPU
    damorgan Oracle ACE Director
    Currently Being Moderated
    Thank you for providing version number and a clear descriptions but I still have an issue understanding what you have. Here are my questions.

    1. If it is RAC there is no such construct as a "standby node" and you say you have two of them. Please be specific ... is this RAC or Data Guard or a combination? Or do you have a three node cluster with all services pointing to only a single server wasting all of the resources of the other three?

    2. Two processes can not be at 100% of CPU any more than I can be 200% awake. Please show us how you arrived at this conclusion, on what hardware and operating system, and include a formatted (read the FAQ to learn how) extract showing what you are seeing.

    3. What processes? Name them and again run a SQL statement and post the output so we can see what you are looking at.

    4. Is anything slow? Are there any problems with the system or are you only reacting to two numbers you think are too high?
  • 2. Re: Sessions with Event "class slave wait" taking 100% CPU
    660273 Newbie
    Currently Being Moderated
    Hi Pal, thanks for your reply!
    Here are the answers:
    1. If it is RAC there is no such construct as a "standby node" and you say you have two of them. Please be specific ... is this RAC or Data Guard or a combination? Or do you have a three node cluster with all services pointing to only a single server wasting all of the resources of the other three?
    A: This is a combination of them. We have 2 nodes in a RAC environment shipping logs to two physical standby databases in another site.


    2. Two processes can not be at 100% of CPU any more than I can be 200% awake. Please show us how you arrived at this conclusion, on what hardware and operating system, and include a formatted (read the FAQ to learn how) extract showing what you are seeing.
    A: top in Linux shows two oracle processes consuming 100% (or nearly of it) of 1 CPU each. This is a 16 cores server (each primary node) so 2 of these are working on full capacity to service the described processes.


    3. What processes? Name them and again run a SQL statement and post the output so we can see what you are looking at.
    A: Like I said before I failed to find which background process they belongs to as there are no indications of them in v$bgprocess and in v$session/v$process either. I can see them in v$process and v$session but again, there are no indications of for which bg process they belong to.


    4. Is anything slow? Are there any problems with the system or are you only reacting to two numbers you think are too high?
    A: No, system is not slow because of that but the client wants to investigate and have it sorted as it is not normal.


    Additionally I found a bug in metalink which seems to be the culprit, but I'm unsure yet. This is Bug 12929268 : HIGH CPU ON ORA_O00N PROCESS

    Is any other information that I may provide you?

    Thanks for help!

    Lauro Ojeda

Legend

  • Correct Answers - 10 points
  • Helpful Answers - 5 points