pcp
[Top] [All Lists]

pcp updates - check/daily issue is a doozie

To: pcp@xxxxxxxxxxx
Subject: pcp updates - check/daily issue is a doozie
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Sat, 07 Sep 2013 06:40:50 +1000
Delivered-to: pcp@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130803 Thunderbird/17.0.8
Changes committed to git://oss.sgi.com/kenj/pcp.git dev

 qa/366                         |   48 +++++++++++++++++++++++++++++++++++++----
 qa/366.darwin.1                |   18 ---------------
 qa/366.darwin.2                |   24 --------------------
 qa/366.linux.1                 |   12 ----------
 qa/366.linux.2                 |   18 ---------------
 qa/366.solaris.1               |   12 ----------
 qa/366.solaris.2               |   18 ---------------
 src/pmie/pmie_check.sh         |    6 ++++-
 src/pmie/pmie_daily.sh         |    6 ++++-
 src/pmlogger/pmlogger_check.sh |   10 +++++---
 src/pmlogger/pmlogger_daily.sh |    6 ++++-
 11 files changed, 66 insertions(+), 112 deletions(-)

commit c7467f70376b8987ccd686436f7ef2937c3ec840
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Sat Sep 7 06:38:20 2013 +1000

    qa/366 - some re-engineering
    
    Provide a more robust way of dealing with pmlogconf evolutionary
    changes and platform dependencies.

commit 623240bd764f8cb8e0803507cd349ee0bb272b4e
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Sat Sep 7 06:30:32 2013 +1000

    pmie/pmlogger control scripts - bad pmproxy handling
    
    Same problem in the "check" and "daily" scripts for pmlogger and
    pmie, namely if the control file contains a host accessed via
    explicit pmproxy, e.g. realhost@proxyhost then badness happens.
    
    Specifically, the scripts fail to recognize that the corresponding
    pmie or pmlogger is running, so by default they will start a
    new one every time the check script runs. For pmlogger, at the
    end of the day the daily script will combine all of the archives
    (with the same data logged from 1 to 48 times!), remove all the
    component archive files _and_ leave all the pmlogger instances
    running _and_ start another one.
    
    And yes, this did happen in a real production environment!

<Prev in Thread] Current Thread [Next in Thread>
  • pcp updates - check/daily issue is a doozie, Ken McDonell <=