.TH knoerred 1 "April 2012" "ngtx"

.SH NAME
knoerred \- udp server for nagios remote checks
.br
knoerrehd \- http server for nagios remote checks
.SH SYNOPSIS
.B knoerre[h]d
.B listenport
.B allowedclientaddress
[
.B allowedclientaddress
] ...
.br
.B /path/to/knoerre[h]d
.B -d
.B listenport
.B allowedclientaddress
[
.B allowedclientaddress
] ...
.br

.SH DESCRIPTION
.B knoerred
is a UDP and
.B knoerrehd
is a http standalone server for nagios remote monitoring.
Like its brother
.B knoerre
it has a small footprint and is very fast.
Unlike
.B knoerre
which is designed to be used with DJB's tcpserver
(or other super-daemons)
.B knoerre[h]d
runs per default as standalone non-forking-daemon with one "slave" thread.
It waits for one packet with a request and
sends one packet back.
The key-request sent to knoerred must be
terminated by a LF.
knoerrehd accepts http requests like knoerre.
.br
.B knoerre[h]d
listens on all local addresses on the specified port.
As a second argument the IP address of a allowed client host
must be specified. Up to 19 optional addresses may be given.


.B knoerrehd
has more limits than
.B knoerre
but it runs as standalone daemon and offers
continous measurement of cpu times and network traffic.
.B knoerred
is also limited but it may run 5 to 20 times faster than
.B knoerre.
.B knoerred
and
.B knoerrehd
do not support calling an external program
or script like
.B knoerre
does.
Currently only a few checks/keys are built in:
.br
.B load1 load5 load15
.B cpuu cpus cpuw cpui cpuI
.B cachedvalue
.B cat
.B cmdline
.B cmp
.B diskinodes
.B diskusage
.B diskusagelocal
.B dmesg
.B filetimestamp
.B mailqsize
.B netlinksdown
.B nettraf0 nettraf1
(aka nettraf/eth0 and nettraf/eth1)
.B proccount proccounttg
.B process processd
.B swaps
.B timediff
.B uptime
.B wc-l
.B loadmulti
.br
Please take look at
.B knoerre(1)
for description of the keys.
.br
.B loadmulti
is a collection of some checks in a single request:
.br
.B cpuI
.B cpus
.B cpuu
.B cpuw
.B ctxt
.B diskinodes/
.B diskusage/
.B hostname
.B load1
.B load15
.B load5
.B netlinksdown
.B nettraf0
.B nettraf1
.B proccount
.B swaps
.B timediff

For "cached" keys like cpu times
.B knoerrehd
only needs 4 syscalls for receive and send.
.B knoerred
only needs 2 syscalls to answer a request.
(only a kernel module would be faster).
If a file
.B /nagios-cache/multiresults
exists then its contents with up to 966 bytes will be included
in the packet sent back.

.SH OPTIONS
.TP
.B \-d
daemonize knoerre[h]d. Remember to specify the full path for this option.

.SH EXAMPLES
.br
What's the load per one minute average on the remote
server running knoerred?
.RS
.br
$ echo load1 |nc -w 1 -u 127.0.0.1 8888
.br
0.00 0.00 0.02 2/250 19914
.br
0.00
.br
$ time check_remote_by_udp -w 7 -c 30 -H 127.0.0.1 -p 8888 -t 1 load1
.br
REMOTE OK - 3.68 | load1=3.68;7;30
.br

.br
real    0m0.002s
.br
user    0m0.000s
.br
sys     0m0.000s
.RE

The following example uses netcat to show the results of a
.B loadmulti
check, including check of hostname and systime:
.RS
$ echo loadmulti/$(date +%s)/gway |nc -w 1 -u 127.0.0.1 8888
.br
 hostname
.br
0
.br
 timediff
.br
0
.br
 cpuu u0s0w0
.br
0
.br
 cpus u0s0w0
.br
0
.br
 cpuw u0s0w0
.br
0
.br
 cpuI u0s0w0
.br
1
.br
 ctxt
.br
70
.br
 nettraf0
.br
0
.br
 nettraf1
.br
0
.br
 swaps
.br
2
.br
 diskusage/ 0GB free
.br
84
.br
 diskinodes/ 47205 free
.br
61
.br
 netlinksdown lo,eth0,eth1
.br
0
.br
 load5
.br
0.00
.br
 load15
.br
0.00
.br
 proccount
.br
138
.br
 load1
.br
0.00 0.00 0.00 3/138 21264
.br
0.00
.RE

.SH CAVEATS
.B knoerred
does only send one single udp packet with up to approx. 1.5 KB
back where all results must fit in.


.SH "SEE ALSO"
.B knoerre(1),
.B check_remote_by_udp_multi(1),
.B check_remote_by_udp(1),
.B check_remote_by_http(1)
.br
http://downloads.tuxad.de/ngtx
.br
http://cr.yp.to/daemontools.html


.SH AUTHOR
Frank Bergmann, http://www.tuxad.com