|
> I'm currently running IPVS 1.0.9 and ipvsadm 1.21 (compiled with popt
> and IPVS 1.0.9). I plan to upgrade before the year is out. All my
> services use LVS-NAT.
>
> I'm writing a Nagios plugin to check details about an LVSed
> service and
> return information to the monitoring host. I started trying
> to get the
> statistics with SNMP, but I couldn't get the LVS-MIB to work
> (quickly).
> Looking at /proc/net/ip_vs_stats, it seems to be overall
> stats for the
> director, not per service. So I thought I'd parse the output of
> 'ipvsadm' to get the statistics.
>
> I'm having a problem, however. The counters for InPkts,
> OutPkts, InBytes
> and OutBytes are often returned by ipvsadm with a K, M, G, or
> T suffix
> (e.g. numbers are in kilobytes, megabytes etc.).
> Unfortunately, this is
> not granular enough for running checks every 5 minutes - the
> counters do
> not update in that time, so it appears there is no traffic
> going through
> for that service.
>
> Can anyone suggest how I might get the stats values in bytes?
> Is there
> something in /proc for IPVS 1.0.9 that I can use, or does a later
> version of IPVS have more detailed stats somewhere?
Neat idea. I use Nagios too, it is really nice.
Have you thought about using rate instead of exact packets?
--rate Output of rate information. The rate information (such as
con-
nections/second, bytes/second and packets/second) of
services
and its servers will be displayed, when the services are
listed.
Regards,
P
|
