pcp
[Top] [All Lists]

pcp2graphite updates

To: pcp developers <pcp@xxxxxxxxxxx>
Subject: pcp2graphite updates
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Thu, 16 Jul 2015 22:28:10 -0400
Delivered-to: pcp@xxxxxxxxxxx
User-agent: Mutt/1.4.2.2i
Hi -

Please see git://sourceware.org/git/pcpfans.git branch fche/pcp2graphite:


commit 9b363d88a744c925860ec2e26ac2e593d804e4d4
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Thu Jul 16 22:18:48 2015 -0400

    pcp2graphite qa/667: test new archive feature
    
    Added subtests 4 & 5, slightly robustified others.

commit 7d428105a9d70d24a69a466724fc7c39f4e5dfce
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Thu Jul 16 22:18:20 2015 -0400

    pcp2graphite improvements: document new -a/-O/-T/-s options

commit 59f5abada48370a78fe492c32a6d46d77b6392af
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Thu Jul 16 22:10:47 2015 -0400

    pcp2graphite improvements: archive mode, reused connections, python3, etc.
    
    RHBZ1242622: support archive relay mode by addition of -a,
    -O, -T, and -s options.  To make this performant, reuse
    a long-lived tcp socket to the carbon-cache server rather
    than drop every message.  (It turns out that droppage was
    not necessary for live mode either.)
    
    Improve diagnostics by intercepting any -Dfoo option to set
    an internal self.debug flag.  -D0 is especially handy, since
    it leaves the libpcp pmDebug flag at zero.
    
    RHBZ1243123: python3 compatibility corrected by dropping the
    .write(... flush=true).
    
    python3 actually function tested against live carbon-cache (rather
    than pcpqa socat stand-in) to confirm actual end-to-end transmission.
    It was necessary to force a downgrade to the pickling format to avoid
    the possibility of a python2 carbon-cache failing to parse a python3
    pickle.

commit 732de928d2fde25e1db227a5521cfc0172d83d5f
Author: Frank Ch. Eigler <fche@xxxxxxxxxx>
Date:   Thu Jul 16 21:07:54 2015 -0400

    pcp2graphite qa/667: make more deterministic
    
    With socat configured to fork child processes, and maintaining plain
    stdout redirection to a file, it's possible for the files to be
    written-to concurrently and thus to lose lines.  We switch to >>
    append-mode, which precludes that accident.

<Prev in Thread] Current Thread [Next in Thread>
  • pcp2graphite updates, Frank Ch. Eigler <=