Oesor's questions

The IETF appears to have had a draft to specify a null mx record whereby a domain would not handle mail and mail delivery systems would fail and return a undeliverable system immediately by directing a domain's only MX record to '.' (c/f https://datatracker.ietf.org/doc/html/draft-delany-nullmx-00)

Is this draft specification followed by most mail servers out there and worth setting up? Alternately, what's the best way to configure mail to not be delivered for a domain in the process of deprication?

Given the following network, with all sites' network routing tables built via OSPF:

   Site_A
      |
Site_A_router
  |           \
  |      Intermediate_site_C_router -- Other_sites
  |                       |
  |      Intermediate_site_D_router
  |                       |
Site_B_router_1     Site_B_router_2
             \       /
            Site_B_LAN
                 |
              Site_B
                 |
          Site_B_router_3
                 |
             Internet

Currently, there is no link between Site A and Site C. I'm looking at adding that in place and seting up unequal cost load balancing between Site A and Site B to provide some redundancy as well as get more effective bandwidth between Site A and hosts on the Internet and at Site B. From a significant amount of Googling, it appears I want to set up MPLS TE tunnels between the Site A router and one of the Site B routers to direct all traffic between the two sites over one of the two routes.

I'm stuck at trying to wrap my head around how the MPLS tunnels will fit into the scheme of the routed network. From my reading, MPLS TE layers on top of the OSPF derived routes and simply defines paths to take between two routers.

Do I need to advertise their existence at all via OSPF? I wouldn't think so, as I suspect that'd just end up routing all traffic from anywhere on the network destined to Site A through Site B then back across one of the two paths to Site A.

Alternately, would I just set the tunnels up on a Site B router but not advertise their existence and just allow them to sort any trafifc that arrives into one of the two tunnels paths? If this works in this manner, I'd expect hosts in site D, site C, and the other sites connected via site C to just route the traffic over the shortest OSPF route, ignoring the tunnels.

Complicating this is that many of the hosts in Site B build their routing tables via OSPF and thus there is no single gateway router that the hosts use to determine the next hop. Which of the three routers should I be setting the Site B -> Site A direction of the MPLS TE tunnel up on in order to get the appropriate load balancing?

If I set it up on router 3, I suspect only Internet-sourced traffic would be load balanced, as hosts in site B are going to be sending site A-destined packets through to either router 1 or 2 directly. Alternately, I could set it up on router 1, as that's almost certain to be the OSPF shortest route. However, then traffic is going to be routed from router 1 through to router 2 somewhat needlessly as traffic takes the alternate tunnel. I suspect this is overthinking things somewhat and is propably premature optimization.

Given this situation, what's the best practice way for doing what I'm trying to do?

Yes, I know about ARP but that relies somewhat on IP layer connectivity in that the remote system must be configured with an IP on a recognized subnet. Is there any way to just dump off what mac addresses the interface has seen Ethernet frames from; either recently or since the interface was initialized -- I'm not picky which.

I'm having some issues with escalations happening properly and I'm not sure if it's because of my config or because the nagios binary is nonstandard and something may be broken. I've got little experience with nagios, and just want to make sure this is being set appropriately.

Should the following config file definition allow the escalations to take over and increment the notification interval as expected? Is there somewhere else in the config files I should be looking at to figure out what's going on? I've enabled debug 32 in the config and it's simply spitting out 'Host notification will NOT be escalated.' for each notification.

The configuration does pass the pre flight check with no issues, and reports that it's parsing the three host escalations in the config.

# test host definition
define host {
    host_name   test
    alias       test
    address     10.0.0.10
    hostgroups  test
    check_interval  0
    retry_interval  1
    max_check_attempts 2
    flap_detection_enabled  0
    icon_image  windows.png
    icon_image_alt  LOGO - Windows
    vrml_image  windows.png
    statusmap_image windows.png
    action_url  /info/host/275
    check_period    24x7
    contact_groups  hostgroup15_servicegroup1,hostgroup15_servicegroup10,hostgroup15_servicegroup13,hostgroup15_servicegroup14,hostgroup15_servicegroup2,hostgroup15_servicegroup3,hostgroup15_servicegroup4,hostgroup15_servicegroup42,hostgroup15_servicegroup45,hostgroup15_servicegroup46,hostgroup15_servicegroup47,hostgroup15_servicegroup5,hostgroup15_servicegroup8,hostgroup15_servicegroup9,ov_monitored_by_master
    check_command       check_host_15!-H $HOSTADDRESS$ -t 3 -w 500.0,80% -c 1000.0,100%
    parents nagios
    notifications_enabled   1
    notification_interval   3
    notification_period 24x7
    notification_options    u,d,r
    use host-global
}

define hostescalation{
    host_name               test
    first_notification      3
    last_notification       4
    notification_interval   10
    contact_groups          hostgroup15_servicegroup1,hostgroup15_servicegroup10,hostgroup15_servicegroup13,hostgroup15_servicegroup14,hostgroup15_servicegroup2,hostgroup15_servicegroup3,hostgroup15_servicegroup4,hostgroup15_servicegroup42,hostgroup15_servicegroup45,hostgroup15_servicegroup46,hostgroup15_servicegroup47,hostgroup15_servicegroup5,hostgroup15_servicegroup8,hostgroup15_servicegroup9,ov_monitored_by_master
}

define hostescalation{
    host_name               test
    first_notification      4
    last_notification       5
    notification_interval   30
    contact_groups          hostgroup15_servicegroup1,hostgroup15_servicegroup10,hostgroup15_servicegroup13,hostgroup15_servicegroup14,hostgroup15_servicegroup2,hostgroup15_servicegroup3,hostgroup15_servicegroup4,hostgroup15_servicegroup42,hostgroup15_servicegroup45,hostgroup15_servicegroup46,hostgroup15_servicegroup47,hostgroup15_servicegroup5,hostgroup15_servicegroup8,hostgroup15_servicegroup9,ov_monitored_by_master
}

define hostescalation{
    host_name               test
    first_notification      5
    last_notification       0
    notification_interval   240
    contact_groups          hostgroup15_servicegroup1,hostgroup15_servicegroup10,hostgroup15_servicegroup13,hostgroup15_servicegroup14,hostgroup15_servicegroup2,hostgroup15_servicegroup3,hostgroup15_servicegroup4,hostgroup15_servicegroup42,hostgroup15_servicegroup45,hostgroup15_servicegroup46,hostgroup15_servicegroup47,hostgroup15_servicegroup5,hostgroup15_servicegroup8,hostgroup15_servicegroup9,ov_monitored_by_master
}

I'm querying some signal level values measured in dBm, and the SNMP host on the remove device reports the values as negative values, ie, -90 dBm. However, check-snmp seems to be incapable of dealing with negative numbers as part of its threshold values. If I specify the values as part of a collection of OIDs, it accepts the syntax but converts the snmp value to positive, thus always generating a WARNING/CRITICAL result:

root@ops-00:/usr/local/nagios/libexec# ./check_snmp -H 192.168.1.100 -o DEVICE-MIB::AverageReceiveSNR.0,DEVICE-MIB::CurrentNoiseFloor.0 -w 10:,~:-85 -c 15:,~:-80 -vvvv
/usr/bin/snmpget -t 1 -r 5 -m ALL -v 1 [authpriv] 192.168.1.100:161 DEVICE-MIB::AverageReceiveSNR.0 DEVICE-MIB::CurrentNoiseFloor.0
DEVICE-MIB::AverageReceiveSNR.0 = INTEGER: 25
DEVICE-MIB::CurrentNoiseFloor.0 = INTEGER: -97
Processing line 1
  oidname: DEVICE-MIB::AverageReceiveSNR.0
  response:  = INTEGER: 25
Processing line 2
  oidname: DEVICE-MIB::CurrentNoiseFloor.0
  response:  = INTEGER: -97
SNMP CRITICAL - 25 *97* | DEVICE-MIB::AverageReceiveSNR.0=25 DEVICE-MIB::CurrentNoiseFloor.0=97

If I run it with a single OID, it gives me an error that the format is incorrect:

root@ops-00:/usr/local/nagios/libexec# ./check_snmp -H 192.168.1.100 -o DEVICE-MIB::CurrentNoiseFloor.0 -w ~:-85 -c ~:-80 -vvvv
Range format incorrect

edit: if I use -w=range it parses properly but still converts the value to positive.

root@ops-00:/usr/local/nagios/libexec# ./check_snmp -H 192.168.1.100 -o DEVICE-MIB::CurrentNoiseFloor.0 -w=~:-85 -c=~:-80 -vvvv
/usr/bin/snmpget -t 1 -r 5 -m ALL -v 1 [authpriv] 192.168.1.100:161 DEVICE-MIB::CurrentNoiseFloor.0
DEVICE-MIB::CurrentNoiseFloor.0 = INTEGER: -97
Processing line 1
  oidname: DEVICE-MIB::CurrentNoiseFloor.0
  response:  = INTEGER: -97
SNMP CRITICAL - *97* | DEVICE-MIB::CurrentNoiseFloor.0=97  

And if I run it with no thresholds defined, it works properly and returns the right value. This makes the graphs correct, however it'll never generate a notification when out of range:

root@ops-00:/usr/local/nagios/libexec# ./check_snmp -H 192.168.1.100 -o DEVICE-MIB::CurrentNoiseFloor.0 -vvvv
/usr/bin/snmpget -t 1 -r 5 -m ALL -v 1 [authpriv] 192.168.1.100:161 DEVICE-MIB::CurrentNoiseFloor.0
DEVICE-MIB::CurrentNoiseFloor.0 = INTEGER: -97
Processing line 1
  oidname: DEVICE-MIB::CurrentNoiseFloor.0
  response:  = INTEGER: -97
SNMP OK - -97 | DEVICE-MIB::CurrentNoiseFloor.0=-97  

What am I doing wrong here? How would I, for example, generate a CRITICAL when the noise floor is -80 dBm or higher, a WARNING when it's -85 to -80 dBm, and an OK when -85 dBm or lower? Do I have to write my own SNMP plugins when dealing with negative values?

I haven't needed anything like this before but I'm sure something exists. I'm looking for a compiled Windows executable (i.e. noninterpreted language) that I don't need to write that can read the arguments passed in on the command line, open an TCP network socket to another machine, and pass the arguments along.

Has anyone used something like this before?

re: netcat for windows: Yeah, basically. Problem is I'm trying to get data out of a WhatsUp installation and onto another box as events are triggered. I can't specify a full command line, just an executable and as a separate field the arguments to pass in, so I can't do any pipe tricks.

AFAIK netcat can only read data from STDIN or a pipe, but there's no way to specify the data as a command line argument. man page seems to support this. Am I mistaken?