Advanced

Monitoring with API

A Presenta*on for MUM Sydney, 2012 By Herry Darmawan

About ME
Herry Darmawan
Working for : Spectrum Indonesia Posi*on : Technical and Opera*onal Manager Home base : Surabaya, Indonesia

Has been using MikroTik since 2004 Daily Ac*vity

Train people how to use MikroTik through MikroTik Cer*ed Training (basic and advance class) Managing technical team of ISP in Surabaya for the last mile connec*on (Wireless and Fiber) Conduc*ng Networking Project and Consulta*on Developing Monitoring and System for network and standard procedure par*cularly using MikroTik as the object

What is This Presentation About?

Monitoring devices What regular method (non-API) cannot achieve? How to use API in polling-based method Case Study !

Regular Monitoring System

Method
ICMP SNMP and SNMP-Trap TCP/UDP checked (based on port)

How to monitor case like this???

10.10.10.100/24 OSPF Network within same Area

Introducing Nagios
Web based monitoring system Modular
Check plugin (in perl or c++) Lots of improvement module (front-end, polling, 3rd party integra*on, etc)

Database backend
NDOMY MySQL Postgres SQL

Recommended Front-End : CENTREON

Centreon host and service

Centreon host details conBig

Centreon plugins for service

Plugin short-name

The real command-prompt syntax (including the parameters)

Centreon - command
The actual command prompt with some MACROs

[root@localhost plugins]# ./check_centreon_snmp_up*me -H 192.168.3.1 -C dsnmp -v 2c d OK - Up*me (in day): 0|up*me=0day(s)

Centreon - attaching to service

Centreon service result

Nagios Plugin Structure

Plugins can be created using perl or c++ (compiled or not) For un-compiled script, this is the structure
Header
Parameter ini*aliza*on Help menu

Process
Processing and gathering informa*on from device

Return Value
Result display RRD result Service status return

Nagios Plugin Structure

Header
Taking parameters from the command prompt Check whether the parameters are correct and complete (for example we need to take the username, but user didnt provide us with the username parameter Print help (if necessary) Global and local variable ini*aliza*on

Process
Is the real process All process (SNMP, Telnet, SSH, API) is happening in this part Beware to check the structure

Nagios Plugin Structure

Return Value
Print a line to send out to Centreon/Nagios as Status Informa*on 2nd line, if any, will be considered as Extended Informa*on Send out a performance data to be graphed using RRD Tool At the end of the script, we have to send out no*ca*on whether this service state is
OK 0 WARNING 1 CRITICAL 2 UNKNOWN 3 or DEPENDENT - 4

Case Study
Scenario
192.168.2.1

MR4
192.168.4.1

MR2 MR1 BGP MR3

OSPF

MR5

192.168.5.1

192.168.1.1

192.168.3.1 192.168.3.2

Monitoring OSPF
What parameter do we need?
Router IP API Port (in this case, we use the default port) Username and Password for the API Interface NAME / NUMBER Threshold Value

We will create a help menu which will be shown if there is uncompleted parameters given

Monitoring OSPF through API

Command Prompt Parameters
usage: $0 m <mtik_ip> -u <user> -p <passwd>! ! -h : help (this message)! -m : hostname or IP of Mikrotik router! -u : admin username! -p : password! -l : list of interface! -i : interface number! -w : warning threshold (in Kbps)! -c : critical threshold (in Kbps)!

Monitoring OSPF through API

Concept
./check_ospf.pl m <RA> -u <U> -p <P> -l! Will list all the corresponding interface inside this router ./check_ospf.pl m <RA> -u <U> -p <P> -i ether1! Will show the OSPF Status, along with the u*liza*on of interface name ether1 with condi*on like this : IF the status of OSPF <> FULL, then considered CRITICAL

About API in PERL

Created by a forum member called cheesegrits He provide some sample source-code and one of it is ac*ng like terminal for API Improvement from the original module :
Accept ? sign rather than only = for the command parameter Improve output (used to be hang for more than 1kB output) Adding some subprocedure
Sub getall_by_key , to list all the result based on .id Sub get_by_key, to get a list of result based on .id as search_key Sub get_by_name, to get a list of result based on custom search_key Sub get_by_value, to get one single value of an item (for example to get the status of interface name ether1)

About API Command

Must be started with Command Word, followed by Auribute Word (or Query Word), then terminated by zero-length Word API Command Word
Its a command in API Almost the same as the terminal command syntax, but no space, instead use / as the replacement Special API command is : getall, login, cancel! Example
/interface/getall! /interface/set! /ip/address/print! /login! /interface/wireless/remove!

About API Attribute

API Auribute Word
Its the value depend on the content of a command Started with = followed by the auribute name, followed by = then end with the auribute value Example
=name=ether1! =status=enable! =.proplist=name,mtu,type,running!

About API Query

API Query Auribute
Used only for print and getall Start with ?, followed by auribute name (or addi*onal command), followed by = then end by auribute value Example
?status (means if THERE IS a auribute named status) ?name=ether1 (means if NAME is ether1) ?-name=ether5 (means if NAME is NOT ether5) ?>comment= (means if there is non-empty comment) ?#<operator> (means popup 2 value just before this query then compare with operator)
The operator can be | (or), & (and), ! change top value with opposite, etc

How to List the OSPF Interface

In terminal, if I want to list the interface, the command is !/interface print! In API, we convert the Terminal Command into API format !/interface/getall! !=.proplist=name!

How to List the OSPF Interface

In PERL, the command will look like this
my(%attrs);! $attrs{'=.proplist'} = name';! my(%results) = Mtik::get_by_key('/interface/getall', \%attrs);! print "List of interface in router $mtik_host\n";! foreach my $item (keys(%results)) {! !my($intno) = $results{$item}{.id};! !my($intname) = $results{$item}{'name'};! !print " $intno - $intname \n";! }!

And the result would be

[root@localhost plugins]# ./check_ospf.pl -m 192.168.3.1 -u api -p test l List of interface in router 192.168.3.1 *3 - ether3 *4 - ether4 *2 - ether2 *1 - ether1

Monitoring OSPF through API

Concept
./check_ospf m <RA> -u <U> -p <P> -l! Will list all the corresponding interface inside this router ./check_ospf.pl m <RA> -u <U> -p <P> -i ether1! Will show the OSPF Status, along with the u*liza*on of interface name ether1 with condi*on like this : IF the status of OSPF <> FULL, then considered CRITICAL

OSPF Neighbor Check

In terminal, the command is !/routing ospf neighbor print! In API, it looks like this /rou*ng/ospf/neighbor/getall ?interface=<interface_name> =.proplist=interface,state,adjacency

OSPF Neighbor Check

#get the interface status based on interface name! $ospfattrs{'=.proplist'} = 'interface,state,adjacency';! my(%results) = Mtik::get_by_name! ! !('/routing/ospf/neighbor/getall', ! ! ! 'interface', $intname, \%ospfattrs);! if (%results) {! # IF the result is non empty, then check the state! !$state = $results{$intname}{'state'};! !$adjacency = $results{$intname}{adjacency};! !if ($state ne "Full) {! ! !$errmsg = "OSPF for $intname status is $state";! ! !$status = "WARNING";! !} else {! ! !$status = "OK";! !}! } else {! # IF the result is empty, then it might be not there! !$errmsg = "OSPF for $intname status not connected";! !$status = "CRITICAL";! } ! !!

Final RESULT
my %ERRORS=('OK'=>0,! ! 'WARNING'=>1,! ! 'CRITICAL'=>2,! ! 'UNKNOWN'=>3,! ! 'DEPENDENT'=>4);!

if ($errmsg) {! !print $errmsg."\n";! } else {! !print "$status : "OSPF status for $intname ! ! ! ! !is $state for $adjacency \n";! }! exit $ERRORS{$status};!

Command Prompt RESULT

### LIST all the interface [root@localhost plugins]# ./check_ospf.pl -m 192.168.3.1 -u api -p test l List of interface in router 192.168.3.1 *3 - ether3 *4 - ether4 *2 - ether2 *1 - ether1 ### RESULT for OK OSPF Status (FULL) [root@localhost plugins]# ./check_ospf.pl -m 192.168.3.1 -u api -p test -i *3 OK : OSPF status for ether3 is Full for 00:43:30 ### RESULT for NOT OK OSPF (status Down or not connected) [root@localhost plugins]# ./check_ospf.pl -m 192.168.3.1 -u api -p test -i *1 OSPF for ether1 status unknown/not connected

Integrate to NAGIOS

$USER1$/check_ospf.pl m $HOSTADDRESS$ -u api p test i $ARG1$! IP Address of the HOST ARGUMENT1 could be dierent for each service /usr/lib/nagios/plugins

Attach it to HOST

Command short-name

ARGUMENT1 : the interface number

TESTING

Drawbacks
API connec*on will constantly ini*ate and closed each *me the monitoring tools doing polling to the device / host Not as fast as SNMP (since we are using TCP Socket conn)

Improvement
Instead of just checking the OSPF status, why dont we check the trac u*liza*on as well and give alert if it reach some threshold?

./check_ospf m <RA> -u <U> -p <P> -i ether1 w 10 c 100! Will show the OSPF Status, along with the u*liza*on of interface name ether1 with condi*on like this : IF the trac u*lized is more than 10kbps (-w 10) then this service status is considered WARNING IF the trac u*lized is more than 100kbps (-c 100) then this service status is considered CRITICAL IF the status of OSPF <> FULL, then considered CRITICAL GRAPH the TX and RX trac

TrafBic Utilization
IF the trac u*lized is more than 10kbps (-w 10) then this service status is considered WARNING IF the trac u*lized is more than 100kbps (-c 100) then this service status is considered CRITICAL First of all, we will take the external value for the WARNING and CRITICAL threshold
WARNING threshold is taken by parameter w! CRITICAL threshold is taken by parameter -c!

TrafBic Utilization
In Terminal we write it like this !/interface monitor-traffic [ether1]! In API, we write it like this !/interface/monitor-traffic! !=once=! !=interface=[ether1]! !

TrafBic Utilization
### TAKING the interface number from the parameter! my($intno) = $options{'i'}; ! ! ### Getting the interface name (the monitor-traffic use name)! $intattrs{'=.proplist'} = 'name';! $intattrs{'.id'} = $intno;! $intname = Mtik::get_value_by_id! ! !('/interface/getall', $intno, 'name', \%intattrs);! ! !! ### Getting the real traffic from monitor-traffic command ! !! $trafficattr{'=.proplist'} = ! ! ! !'rx-bits-per-second, tx-bits-per-second';! $trafficattr{'=once'} = '';! $trafficattr{'=interface'} = $intname;! my(%traffics) = Mtik::get_by_key! ! !(/interface/monitor-traffic, \%trafficattr);! $txbits = $traffics{$intno}{'tx-bits-per-second'};! $rxbits = $traffics{$intno}{'rx-bits-per-second'};!

TrafBic Utilization
Now we compare the bits received with the actual Threshold
if ($txbits > $warningbits || $rxbits > $warningbits) {! !$retmsg .= " but the traffic exceeded the threshold";! !$status = "WARNING";! } elsif ($txbits > $criticalbits || $rxbits > $criticalbits) {! !$retmsg .= " but the traffic exceeded the threshold";! !$status = "CRITICAL";! }! ! print "$status : $retmsg \n";! printf("Traffic Utilization : TX : %.2f ".$txprefix."bps/ ! ! RX : %.2f ".$rxprefix."bps\n! !,$txdispbits,$rxdispbits);! print "|traffic_in=".$txbits."Bits/s;! ! ! !$warningbits;$criticalbits ! ! traffic_out=".$rxbits."Bits/s;! ! ! !$warningbits;$criticalbits\n";! exit $ERRORS{$status};!

TrafBic Utilization - COMMAND

### When the OSPF is OK and the traffic is OK! [root@localhost]# ./check_ospf.pl -m 192.168.3.1 -u api -p test -i *4! OK : OSPF status for ether4 is Full for 00:49:37 ! Traffic Utilization : TX : 0.00 bps/ RX : 0.00 bps! |traffic_in=0Bits/s;100000;1000000 traffic_out=0Bits/s;100000;1000000! ! ! ### When the OSPF is OK but the traffic exceed the threshold! [root@localhost]# ./check_ospf.pl -m 192.168.3.1 -u api -p test -i *3! WARNING : OSPF status for ether3 is Full for 00:01:49 ! but the traffic exceeded the threshold ! Traffic Utilization : TX : 131.97 kbps/ RX : 130.43 kbps! |traffic_in=131968Bits/s;100000;1000000! traffic_out=130432Bits/s;100000;1000000!

Visual Result

Whats NEXT?
Basically we can monitor and graph anything
Graph BGP prexes received and alert when the BGP DOWN or the prexes reach some low threshold Graph the number of Ac*ve Hotspot user, Host that connected to a Hotspot server, and the number of DHCP Lease that has been established Graph the number of sta*on that connect to an Access Point Graph TX/RX Rate and CCQ of a connec*on and send alert once they goes below certain threshold

Centreon and Nagios also provide

Passive Check Lots of Modules and Plugins

 hup://project.spectrumindo.com hup://www.mikro*ktraining.co.id

FURTHER QUESTION
[email protected]