pcp
[Top] [All Lists]

pcp updates: qa, libpcp+pmstat hostnames, pmcc

To: pcp@xxxxxxxxxxx
Subject: pcp updates: qa, libpcp+pmstat hostnames, pmcc
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon, 2 Jun 2014 04:48:19 -0400 (EDT)
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <947661467.18646939.1401698759485.JavaMail.zimbra@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: EJv16SrNRg1tX0nBgXVMIT5hpB0mAg==
Thread-topic: pcp updates: qa, libpcp+pmstat hostnames, pmcc
Changes committed to git://git.performancecopilot.org/pcp.git dev

 qa/297.out               |    2 +
 qa/734                   |   40 +++++++++++++++++++++++++++++++++++++++
 qa/734.out               |    3 ++
 qa/766                   |   34 +++++++++++++++++----------------
 qa/766.out               |    2 +
 qa/776                   |   48 ++++++++++++++++++++++++++++++-----------------
 qa/776.out               |    2 +
 qa/group                 |    1 
 src/libpcp/src/context.c |   47 +++++++++++++++++++++++++++++-----------------
 src/pmstat/pmstat.c      |    9 ++++----
 src/python/pcp/pmcc.py   |    6 +----
 11 files changed, 136 insertions(+), 58 deletions(-)

commit 4ba8fabb8ac792e7cf5fcf2a04bade601b26cf69
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Mon Jun 2 18:44:20 2014 +1000

    Update test 297 to reflect new libpcp context diagnostics

commit 5c8a28e13ac9843edd1f392909e515c77bb96f32
Author: Mark Goodwin <mgoodwin@xxxxxxxxxx>
Date:   Mon Jun 2 15:09:05 2014 +1000

    Fix pmcc updating in-memory result values
    
    An overzealous coding guard was preventing Metric class values
    from being computed correctly from fetched pmResult structures.

commit 3f0595ab30b6e7c5bc4cf3b65fdc27e300c12102
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Mon Jun 2 15:02:56 2014 +1000

    Fix pmstat hostname reporting when running in cluster mode
    
    Found a couple of bugs in the hostname reporting done by
    pmstat and the pmGetContextHostName routine.  Firstly, in
    pmstat a missing field initialisation in the context data
    structure meant the hostname pointer could point off into
    the weeds.  Secondly, the libpcp code that retrieves and
    uses the pmcd.hostname metric didn't correctly setup the
    context for pmLookupName/pmFetch with the result that the
    original context was re-used and the wrong hostname ended
    up being reported.
    
    Add test qa/734 to exercise the whole horrible situation
    and add some libpcp diagnostics around context switching
    in pmGetContextHostName.

commit b30c9aa51be78294944b4e7775271928ce05a92b
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date:   Mon Jun 2 10:03:09 2014 +1000

    No need to check for pminfo in QA tests, nothing would work

commit d7ed4e206d21c81a6d3b5627efceb02b94df94b5
Author: Dave Brolley <brolley@xxxxxxxxxx>
Date:   Fri May 30 15:09:55 2014 -0400

    Make qa tests 766 and 776 (pmfind with active probing) more robust.
    
    - Stole Nathan's deterministic output fix from tests 775 and 946
    - Use pminfo to obtain the address of an active interface instead
      of ifconfig (also Nathan's suggestion).

<Prev in Thread] Current Thread [Next in Thread>
  • pcp updates: qa, libpcp+pmstat hostnames, pmcc, Nathan Scott <=