rpnabar at gmail.com
Wed Aug 4 12:19:43 CDT 2010
On Wed, Aug 4, 2010 at 12:07 PM, James Bensley <jwbensley at gmail.com> wrote:
> On 4 August 2010 17:52, Rahul Nabar <rpnabar at gmail.com> wrote:
>> What if you monitor in-band using something like pings, heartbeat,
>> ganglia etc.? Then use ipmi (via ipmitool) only when stuff goes wrong
>> and a machine is hung or crashed.
> Not a bad idea but with IPMI waiting till something goes POP is silly,
> I could have already been using it to see the temperature on my CPUs
> rising, or the 2nd power supply flapping etc etc...
I monitor temperatures via lm_sensors. Again in-band. I try to keep my
monitoring in-band unless there is a compelling reason to use ipmi.
Maybe some sensors are not available to lm_sensors.
Of course, there has to be some aggregation tool whenever you have
many servers. But that's a different issue from whether to use in-band
(lm_sensors) or out-of-band (ipmi). Personally, both nagios and
ganglia have worked well for the aggregation and display.
More information about the Linux-PowerEdge