pcp
[Top] [All Lists]

pcp updates

To: pcp@xxxxxxxxxxx
Subject: pcp updates
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 14 Jan 2010 20:24:20 +1100
Reply-to: kenj@xxxxxxxxxxxxxxxx
The dbpmda one fixes Nathan's problem reported yesterday.

The others are the tail end of the derived metrics changes to address
some corner cases that are unlikely to be seen with well-behaved PMDAs.

Changes committed to git://oss.sgi.com/kenj/pcp.git dev

 src/dbpmda/src/gram.y         |   45 +++++
 src/libpcp/src/derive.c       |   54 +++---
 src/libpcp/src/derive.h       |   26 +--
 src/libpcp/src/derive_fetch.c |  334 +++++++++++++++---------------------------
 src/libpcp/src/pmns.c         |   10 +
 src/pmdas/sample/help         |   18 ++
 src/pmdas/sample/pmns         |   19 ++
 src/pmdas/sample/src/sample.c |  315 ++++++++++++++++++++++++++-------------
 8 files changed, 453 insertions(+), 368 deletions(-)

commit 29b0c31dad838d8f1f485b5749c74706a8f409e1
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Thu Jan 14 20:18:35 2010 +1100

    dbpmda fix for name lookup and dynamic metrics
    
    As reported by Nathan, this fixes a problem where dbpmda is given
    a metric name, and the translation to a pmid does not follow the
    pmcd-like behaviour and call on the pmda for assistance when the
    name is within a pmns subtree rooted at a dynamic metric.

commit 7174480b8dc5d24acd3e6fa0116b51c602811c99
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Thu Jan 7 06:51:43 2010 +1100

    pmLookupName - refine error handling with derived metrics
    
    The return status is a little tricky as we make a second attempt for
    a derived metric lookup for each unknown name if the original lookup
    fails for any metric name, leading to a possible multiplicity of error
    statuses ... prefer the status from receive_names(), unless all of the
    remaining unknown names are resolved by __dmgetpmid() in which case
    success (numpmid) is the right return status.

commit 40468e684637f8ab24f4da84d111163ac79b2462
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Tue Jan 5 08:03:03 2010 +1100

    sample PMDA - add sample.scramble.*
    
    Some strange metrics with ill-behaved instance domains and in which
    instances come and go, and the order of the returned instances is
    random ... all based on the same indom as sample.bin
    
    Also added a reset button (sample.scramble.version) to make the behaviour
    repeatable for QA via pmstore.

commit c8f9ef4e2aee94c9b812b1d815d0c1f9b9c161a3
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Tue Jan 5 08:01:09 2010 +1100

    Derived metrics - minor fix up for fetch
    
    Correct corner cases when instances are missing, added and the
    order of instances is random between consecutive pmFetch calls.
    Also added some more verbose diags around this code.
    
    All triggered by the bizarre sample.scramble metrics recently
    added to the sample PMDA.

commit 44e7728d3b34ad185c4a924711e6d994b1e0966e
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date:   Mon Jan 4 08:06:34 2010 +1100

    Derived metrics - refinement
    
    After gcov analysis and some expanded coverage in the QA suite, there
    is some dead code removal here, and additional metrics in the sample
    PMDA.


<Prev in Thread] Current Thread [Next in Thread>