[collectd-changes] collectd, the system statistics collection daemon: Changes to 'master'

[Florian Forster]

 configure.in   |    2 +-
 src/collectd.h |   30 ++++++++++++++++++++++++++++++
 src/disk.c     |    2 +-
 src/network.c  |   10 +++++++++-
 src/plugin.h   |    4 +++-
 src/wireless.c |    8 ++++++--
 6 files changed, 50 insertions(+), 6 deletions(-)

New commits:
commit 430b0b4d0275f7fea64376fc166e81fc0bcc36bf
Merge: 1d6a83e9758aeba9f8cebac1d27bd94e46a83450 c60c6dfa5960b23bdc59341ae73e701dd59d4305
Author: Florian Forster <octo at leeloo.lan.home.verplant.org>
Date:   Tue Jun 10 17:10:05 2008 +0200

    Merge branch 'collectd-4.4'

commit c60c6dfa5960b23bdc59341ae73e701dd59d4305
Merge: 889ca6f42e5b2f567ea038a35f8e82209edc0eeb e8c6efb8bf77e835c28ff2c14f13aecbb73e3e28
Author: Florian Forster <octo at leeloo.lan.home.verplant.org>
Date:   Tue Jun 10 17:09:46 2008 +0200

    Merge branch 'collectd-4.3' into collectd-4.4

commit e8c6efb8bf77e835c28ff2c14f13aecbb73e3e28
Author: JoW <xm at leipzig.freifunk.net>
Date:   Mon Jun 9 13:27:16 2008 +0200

    wireless plugins: Interpret noise/power values >100 as (dBm + 256).
    
    Hi list,
    
    I currently develop an embedded statistics application and web interface
    for OpenWrt Kamikaze based on Collectd and RRDTool.
    
    While working with the wireless plugin I noticed some incompatibilities
    with different drivers, namely the Broadcom proprietary driver and madwifi.
    
      /proc/net/wireless on Broadcom:
    
       Inter-| sta-|   Quality        |   Discarded packets               | Missed | WE
        face | tus | link level noise |  nwid  crypt   frag  retry   misc | beacon | 18
          wl0: 0000    2.  179.  163.       0      0   8040    723      0        0
    
      With the Broadcom proprietary driver you get the actual signal and noise
      dbm units by substracting 255 from the values in /proc:
    
       level: 179 - 255 = -76 dBm
       noise: 163 - 255 = -92 dBm
    
      /proc/net/wireless on Atheros SoC:
    
       Inter-| sta-|   Quality        |   Discarded packets               | Missed | WE
        face | tus | link level noise |  nwid  crypt   frag  retry   misc | beacon | 22
         ath0: 0004   21.  -73.  -94.   21841      0      0      0      0        0
    
      Madwifi writes the actual dBm values into /proc.
    
    So we have (at least for now) three possible types of values:
    
       ( x > 0.0   && x <= 100.0 )	current implementation: x is in percent (?)
    
       ( x > 100.0 && x <= 255.0 )	broadcom: range intersects with percents but
    				values below 100.0 (100.0 - 255.0 = -155.0 dBm)
    				are somewhat unlikely to occur
    
       ( x < 0.0 )			atheros: value is already in dBm
    
    A patch which implements the two additional possibilities in collectd's wireless
    plugin can be found in the OpenWrt Trac:
    
    https://dev.openwrt.org/browser/packages/utils/collectd/patches/200-wireless-compat.patch
    
    The patch was made against v4.4.0 of collectd but should work with v4.4.1 too.
    It would be nice to have this in the next release or so :)
    
    Greetings,
    JoW

commit 8c086dac4095529e02530b28fd92e50ad296acce
Author: Alessandro Iurlano <alessandro.iurlano at gmail.com>
Date:   Sat Jun 7 15:16:18 2008 +0200

    disk plugin: Use correct name under Linux 2.4
    
    Hello everybody.
    
    I have noticed that disk graphs are displayed by collection.cgi as
    disk (71687000) instead of the correct disk name on 2.4.x linux
    kernels.
    The problem seems caused by the fact that the position of disk name of
    /proc/partitions (used in 2.4 kernels) differs from /proc/diskstats
    (used in 2.6 kernels).
    
    I include as attachment a patch that seems to fix this problem.
    I have tested it on 2.4.20 (REDHAT 9) and 2.6.24.4 (Debian Sid) systems.
    
    Thanks a lot for Collectd. It is a _very_ useful program.
    
    Alessandro
    
    Signed-off-by: Florian Forster <octo at huhu.verplant.org>

commit 48b794525072b1ff00a90497ef0bb1f4030510f4
Author: Florian Forster <octo at leeloo.lan.home.verplant.org>
Date:   Sat Jun 7 20:39:04 2008 +0200

    network plugin: Don't blindly re-initialize everything in the init callback.
    
    Under Solaris the init functions are called every time the kstat chain
    is updated. Since, among other things, send_buffer_fill was set to zero
    here, the plugin ``forgot'' about values.
    
    Thank you very much to Eric LeBlanc for debugging this issue.

commit 136326a3ccded5e4f2944c38825f2a21d11aa58e
Author: Florian Forster <octo at leeloo.lan.home.verplant.org>
Date:   Fri Jun 6 12:32:13 2008 +0200

    src/collectd.h: Try even harder to determine endianess.
    
    NexentaStor 1.0.2 doesn't have <endian.h>, for example :/

commit b886f9aaf7675f1f9f242a2b9dfe119ce0672d66
Author: Florian Forster <octo at leeloo.lan.home.verplant.org>
Date:   Fri Jun 6 12:21:01 2008 +0200

    src/plugin.h: Use __attribute__((format)) for plugin_log.
    
    All __attribute__'s will be automatically set to /**/ when not using the
    GNU CC via a macro in src/collectd.h. This way this should be portable.