[HPADM] SUMMARY: strange system bottleneck during Oracle 7.3.4 startup on HPUX 10.20 K580

Ted.Fisher_at_dana.com
Date: 06/30/03

  • Next message: Quy NGUYEN DAI: "[HPADM] HP DesignJet configuration"
    To: hpux-admin@dutchworks.nl
    Date: Mon, 30 Jun 2003 10:11:29 -0400
    
    

    I got several replies after I vented about receiving email from vendors
    through a technical
    resource list which you can tell I am not fond of. Thanks to Paveza, Gary, Johnson, Craig E,
    Aaron Bennett, Medaglia, Chris, IlicR@stentor.ca, Abramson, Stuart, Jack
    Gallagher, Denny
    Koovakattu, Alex Vinson for the responses as well as to Rita Workman for her original response.

    Most suggesting watching memory, swap, kernel parms, and system as well as
    Oracle
    logs (which say nothing during the hang period). I've reviewed through
    those items and virtually
    nothing comes up to indicate what might be the bottleneck. SGA sizes were
    mentioned by many
    and these two instance each have SGA slightly less than 500MB. Since
    there is little else
    consuming shared memory we are still below the 1 GB ceiling. A couple of
    people suggested
    checking into Oracle patches that address some similar issues.
    Specifically level 7.3.4.4
    was suggested. I think we'll try to get the application people to try
    this route. The most unusual
    thing about this was that something the DB was doing was holding up other
    user processes, like
    my glance session, for long periods of time. I've never seen that occur
    except when there is a
    major IO hang (like a dead disk). Yet, systems logs and STM did not
    indicate any hardware
    failures. Plus this apparently has been happening every time the second
    instance is brought up.
    Seems very strange to me

    Thanks for the help.

    Ted F. Fisher
    Dana Corporation
    Global Information Technology Group
    Technical Team
    Ted.Fisher@dana.com

    My orginal post(s):

    Before I vent I'd like to thank Rita Workman who was the only person to send me a real reply. She suggested looking at
    Oracle patch levels which I will pursue.

    Unfortunately this list has apparently deteriorated from the helpful
    technical resource it once was because two out of three responses I
    received came from hardware vendors hoping to get some business. Responses
    like the one below are not helpful or desired! The only affect it will
    have is that I will NEVER buy hardware from them for having abused a
    resource that was once helpful to those looking to pool technical knowledge. It is unfortunate that a few individuals can disrupt the
    value of this list and don't realize that they are hurting - not helping -
    any chance that we might buy from them. I'll not use the list any more
    unless I have a dire emergency because it's not worth putting up with the
    headache. Sorry for ranting.

           We have a K580 running HPUX 10.20 and two Oracle 7.3.4 SIDs. When
    the second SID is started it appears to hang the entire system. Until the
    second SID is up the first is completely frozen (user sessions), glance
    will not start or update if already running, etc. The load average climbs
    to over 12 while the DB is starting and drops once it's up. Top will run
    and shows the CPU very low which would indicate that the high load value
    is due to IO. Sar shows disk queues pretty low though (=< .50). I can
    log in a new session with no problem. The behavior of the system is
    similar to when a disk dies and IO's get hung causing any IO related
    process to hang. But, I can't find what kind of IO is hanging or why. We
    have other similar systems which don't appear to have this problem. Has
    anyone seen this kind of behavior before or have any suggestions as to how
    t! o identify the bottleneck?!

    --
                 ---> Please post QUESTIONS and SUMMARIES only!! <---
            To subscribe/unsubscribe to this list, contact majordomo@dutchworks.nl
           Name: hpux-admin@dutchworks.nl     Owner: owner-hpux-admin@dutchworks.nl
     
     Archives:  ftp.dutchworks.nl:/pub/digests/hpux-admin       (FTP, browse only)
                http://www.dutchworks.nl/htbin/hpsysadmin   (Web, browse & search)
    

  • Next message: Quy NGUYEN DAI: "[HPADM] HP DesignJet configuration"

    Relevant Pages

    • Re: Mapping Networked Printers Using Command Prompt
      ... There are a lot of reasons why you might not see a resource in Printers and ... Faxes (which is a subset of My Network Places). ... answer below my responses: ...
      (microsoft.public.windowsxp.network_web)
    • Re: Mapping Networked Printers Using Command Prompt
      ... There are a lot of reasons why you might not see a resource in Printers and ... Faxes (which is a subset of My Network Places). ... answer below my responses: ...
      (microsoft.public.windowsxp.network_web)
    • Re: Forwarding to next php script on correct input?
      ... For 201 responses, the Location ... is that of the new resource which was created by the request. ... the location SHOULD indicate the server's preferred URI for ...
      (comp.lang.php)
    • Re: Changing Resource % mid project.
      ... you need to make sure it is even *possible* to arbitrarily change the resource allocation like you want to show in the plan. ... You mention that these "tasks" involve waiting for responses. ... If you show it as a task assigned to a resource it means work is being performed on it consuming the resource's energy and the cost of that work is included in your budget. ...
      (microsoft.public.project)
    • Re: Actual work from unopened projects not reflected in resource u
      ... suggesting approaches, could you explain to me why it is important to have ... > though for a given resource the number of files is quite a bit smaller. ... i.e. using a resource pool without server. ... >>> Suppose I want to use load levelling. ...
      (microsoft.public.project)