pcp
[Top] [All Lists]

pcp updates: pmmgr bugfixes,qa

To: pcp developers <pcp@xxxxxxxxxxx>
Subject: pcp updates: pmmgr bugfixes,qa
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Fri, 16 Jan 2015 21:53:19 -0500
Delivered-to: pcp@xxxxxxxxxxx
User-agent: Mutt/1.4.2.2i
Hi -

Please see pcpfans.git branch fche/pmwebd for some important bug
fixes, and another attempt to get a pmwebd test case into pcpqa.


commit 368a646674b5d775dbebe1ea33ab846ba0b41054 (HEAD, origin/fche/pmmgr, 
fche/pmmgr)
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Fri Jan 16 21:47:02 2015 -0500

    pmmgr: align poll time with clock intervals
    
    Because polling, and subsequent handling of dead/live hosts, the real
    intervals between polls can drift.  Counteract this by aligning sleep
    amounts to wake up at aligned intervals (e.g., at :00 seconds for
    1-minute poll intervals).  Test case 666 updated with more explicit
    pmcd starts/stops to make output more regular.

commit 7029504b12cd8c25e64cb85807d90fb278524e73
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Fri Jan 16 18:44:39 2015 -0500

    pmwebd qa/666: now only four minutes
    
    This old test case is now shrunk down to apprx. 4 minutes' runtime.

commit 90ea027d31629dd74278929d7c8b69f2f72b522b
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Fri Jan 16 09:32:05 2015 -0500

    pmmgr fix: handle rapidly-changing ip addresses
    
    It can happen that a pmmgr-monitored target changes its ip address so
    quickly that it never appears 'down' between two poll intervals.  In
    this case, pmmgr didn't treat the renumbered target as requiring
    daemon stopping/restarting, and instead futilely tried to restarting
    the daemons against the old address.
    
    This is corrected by checking the new preferred-address for a host
    against the previously preferred one, and treating a change as a
    dead/alive transition.
    
    Testing included multiple VMs on a network hand-changing addresses
    via avahi and/or raw ip, to stress both libpcp discovery streams.
    (Mechanising this test is unfortunately not obvious.)

commit f29b3a63c05a2361e924bf04c7dc426b8f0e1234
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Thu Jan 15 14:24:46 2015 -0500

    pmmgr: rpm packaging tweaks
    
    Because pmmgr is configured by individual files, where their
    non-existence can be a signal (such as to disable pmie).  This is a
    problem if a later package (RPM) upgrade brings a new copy of
    /etc/pcp/pmmgr, including the deliberately rm'd files.  To fix this,
    rpm allows such config files to be marked %config(missingok,noreplace),
    which is henceforth applied to the pmmgr ones.  While in the vicinity,
    toned down the cautionary language used to describe the subrpm.


<Prev in Thread] Current Thread [Next in Thread>
  • pcp updates: pmmgr bugfixes,qa, Frank Ch. Eigler <=