[Nagiosplug-help] check_icmp seems flapping - followon to RE: make of nagios-plugins-1.4.5 on AIX 4.3 fails

[Ralph.Grothe at itdz-berlin.de]

I am still seem to have some serious trouble with my build of the
check_icmp plugin.

Because the make was prematurely aborted (owe to the check_swap
error)
I manually chown-ed root of check_icmp and chmod-ed u+s 
because ICMP packet generation I assume requires root privileges.

I then copied it into $USER1$ and there set a hard link to
check_host.

In my hosts.cfg template I defined a check-host-alive as default
check_command
that looks like this

define command {
    command_name	check-host-alive
    command_line	$USER1$/check_host -H $HOSTADDRESS$ -t 15
-c 10000
}


After a bit of further tweaking of my config files to reflect a
hopefully cleaner overall layout
I uncautiously started the new 2.5 nagios after all pre-flight
checks were satisfied
without prior disabling of host notifications.

I then was shocked to realize that nagios was cheerfully churning
out dozens of alert notification
when the hosts' states changed from soft critical to hard
critical.
Only to minutes later relaps from hard critical to hard ok, and
notifying about the recovery
(because host notification_options of course included r in my
template).
This was kind of flip flopping for many hosts.

I then ran check_host several times manually where I realized the
following hanging:


$ ~/libexec/check_host -H somehost
mode: 1
CRITICAL - somehost: rta nan, lost
100%|rta=0.000ms;1000.000;1000.000;0; pl=100%;
100;100;; 


But an instantly followed ping always returned the echo requests:
 
$ ping -c 3 somehost
PING somehost.somewhere.tld: (123.123.123.123): 56 data bytes
64 bytes from 123.123.123.123: icmp_seq=0 ttl=248 time=3 ms
64 bytes from 123.123.123.123: icmp_seq=1 ttl=248 time=3 ms
64 bytes from 123.123.123.123: icmp_seq=2 ttl=248 time=3 ms

----somehost.somewhere.tld PING Statistics----
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max = 3/3/3 ms


Now I am curiuous whether my compilation of check_icmp is ok?

Of course, could I substitute check_host in my check-host-alive
definition by a check_ping or check_fping
invocation.
But I thought this would defeat the purpose of the check_host
functionality.

Have I missed a crucial step that the Makefile would have taken
care of
if my make all hadn't failed prematurely?


Cheers
Ralph 




> -----Original Message-----
> From: Andreas Ericsson [mailto:ae at op5.se]
> Sent: Wednesday, November 29, 2006 1:47 PM
> To: Grothe, Ralph
> Cc: nagiosplug-help at lists.sourceforge.net
> Subject: Re: [Nagiosplug-help] make of nagios-plugins-1.4.5 on
AIX 4.3
> fails for check_swap with "Undefined symbol: .floor"
> 
> 
> Ralph.Grothe at itdz-berlin.de wrote:
> > But am I right in assuming that all that is required after
the
> > individual make of only the required plugins
> > was to copy them into $USER1$ (aka NAGIOS_ROOT/libexec)?
> > 
> 
> Yes.
> 
> > 
> > Obviously 29 plugins were built successfully from my "make
all"
> > until make check_swap caused the premature abort.
> > 
> > 
> > $ ls -F|grep -c \*\$
> > 29
> > 
> > 
> > But there seem to be many more source files in the plugins
subdir
> > of the source tree
> > 
> 
> Some of those are library files. Others are only built if your
system 
> meets certain preconditions (check_pgsql, check_mysql and the
likes).
> 
> > $ echo *.c|wc -w
> >       44
> > 
> > I can see that at least one vital plugin so far is missing.
> > But I can't find a source file for it.
> > 
> > $ ls -l check_{udp,tcp}.c
> > ls: 0653-341 The file check_udp.c does not exist.
> > -rw-r--r--   1 nagios   nagios     18728 Oct 20 01:53
check_tcp.c
> > 
> > 
> > Could this mean that check_udp is merely a link to check_tcp
(or
> > some other check_*),
> > and that when invoked in this guise the check behaves
accordingly
> > 
> 
> Yes.
> 
> > 
> > I also think to recall to have heard somewhere that
similarily if
> > one links check_host to check_icmp,
> > and uses this name as the general host check, that nagios
would
> > then only send a single ICMP packet
> > to a service's host, when that service check returned
unknown,
> > instead of several like check_icmp does?
> > 
> 
> Moderately correct. When check_icmp is invoked as check_host, 
> it stops 
> sending as soon as it receives any valid ICMP response what 
> so ever. It 
> still sends 5 packets (per default) before giving up 
> entirely. This is a 
> very useful thing to do since it means a hostcheck can be 
> completed in 
> round_trip_time+system_time (usually 10 - 300 milliseconds), 
> instead of 
> always using at least 5 seconds as was the default earlier.
> 
> -- 
> Andreas Ericsson                   andreas.ericsson at op5.se
> OP5 AB                             www.op5.se
> Tel: +46 8-230225                  Fax: +46 8-230231
>