RE: [Vacm-develop] PATCH: vacm_sys_statd and flim enhancements

[San M. <net...@va...>]

Heh.. all and any help is appreciated...

Zac and I are actually preparing a 'statement' as to the future of
vacm... without elaborating ill just say that we plan to keep developing
the new architecture even though VA has decided not to participate...


-----Original Message-----
From: vac...@li...
[mailto:vac...@li...] On Behalf Of Mike
Snitzer
Sent: Tuesday, September 18, 2001 10:51 AM
To: vac...@li...
Subject: [Vacm-develop] PATCH: vacm_sys_statd and flim enhancements


Hello,

I work for Paralogic, Inc and have added some useful features to
vacm_sys_statd and flim.  I have attached the patch (112K) that applies
cleanly to vacm-2.0.5 (patch details listed below), there are also RPMS
and SRPMS at: ftp://ftp.plogic.com/pub/vacm/

I figured the greater vacm community (what's left of it) would
appreciate
these new features, please email me with any feedback/bug
fixes/recommendations you might have.  

I now plan to make systat.p in flim have a threaded poll_timeout() (1
data
collection thread for each node).  This seems like a logical way to
achieve increased scalability during data collection. SAN, will the
nexxus
support multiple simultaneous connections?  At first glance/trial the
nexxus seemed to be ok; please advise.

I also will be adding alarm status monitoring. That is, establish the
desired threshold for a particular statistic; and if exceeded notify the
configured contact with information regarding the alarm.  I'd like to
make
use of vacm::perl coupled with mon and just have flim have a plugin to
facilitate alarm threshold configuration.  ZAC, care to fix vacm::perl
so
that it actually works with vacm-2.0.5?  Again if anyone has other
suggestions please advise.

I also plan on making the flim interface and certain functionality of
its
plugins suck less, and I'd appreciate it if any of you helped me out.

Thanks,
Mike

-----
Paralogic PATCH - overview of changes:

vacm_sys_statd:
  SYSSTAT:SENSORS:<nodename>
  - support for cpu temp and fans data collection via lm_sensors
  (libsensors), supported chips include: w83781d, w83627hf, lm75, lm78,
  lm87 (serverworks), adm1025 (via)
  
  SYSSTAT:NET:<nodename>
  - support for network statistics data collection, provides the ip
  address for each network interface, a data collection time stamp, and
  all data in /proc/net/dev

  SYSSTAT:ECC:<nodename>
  - support for ECC Single and Multi bit errors, bank number, and memory
  size is available if ECC kernel module is loaded (/proc/ram)

  SYSSTAT:FS:<nodename>
  - support for devfs filesystems

  I've also added some configure flags;
  --without-sensors      do not build with lm_sensors support
  --without-plogic       do not build reduced paralogic vacm (use this
  flag to build emp, baytech, msc, va1000, vasenet, sbt2, quanta)

Flim:
  - node state is now monitored via src/node.c:node_state_timeout(), and
  uses flim_plugin_broadcast() to send "node_up" or "node_down" to all
  plugins only when a node changes state, the default timeout for
pinging
  each node is 60 seconds.

  - when a plugin sends the "ready" message back to the client, it now
  gets the state of all nodes (UP or DOWN, on first pass nodes are
  initialized to UP) in addition to what "ready" originally did.

  - icon for node in "Nexxus & Node" tree is now displayed according to
  nodes' availability; red icon for a node that is down, normal 1u node
  icon if up.

  - conf.c: moved nexxus startup to before plugins, makes more sense

  systat.p:
  - Added notebook pages for lm_sensors, ecc, and network usage.  ECC
and
  lm_sensors are only collected on nodes that support them, as
determined
  on the first pass of data collection on all nodes
  
  - Misc now displays Linux Version information
  
  - if a node is not reachable (via icmp_echo:ping:<nodename>) the whole
  frame for <nodename> is hidden, when it becomes available again so
will
  the frame.

  - systat.p.c will react to a node's sys_statd dying by hiding the data
  presentation box; and changing the label to be: "No Sysstat support on
  <node>"
  
  - when a node or sys_statd of a node returns to an UP state, the
  graphical display changes accordingly