One of the most interesting and very often asked things is how do I configure SCOM 2012 as a SNMP trap receiver? The solution isn't really easy and a lot of SNMP basics are needed.
Again Microsoft need to do some homework to make some things much more easier and improve SCOM regarding SNMP traps. But first I would like to thank some people for their good documentation which pushed me to the right direction:
I've done some investigations of our System Center Operations Manager 2012 R2 update rollup 2 installation and here are my findings:
To get more familiar with the SCOM SNMP network device discovery I build following scenario:
Provisioning of a RedHat Enterprise Linux 5 VM and installation of the net-snmp suite. This software package delivered a full featured SNMP daemon (snmpd) and additionally a snmptrap daemon (snmptrapd). All daemons (aka services) are highly configurable and there is very good documentation.
[root@<hostname> ~]# vi /etc/snmp/snmpd.conf ... # groupName securityModel securityName group notConfigGroup v1 notConfigUser #group notConfigGroup v2c notConfigUser #### # Third, create a view for us to let the group have rights to: ...
[root@<hostname> ~]# service snmpd restart Stopping snmpd: [FAILED] Starting snmpd: [ OK ] [root@<hostname> ~]#
If this is all done, don't forget to save and configure SCOM to detect the new network device. Don't discover it as a Linux or Unix computer!
Before performing the discovery I started a port scan with nmap and found that only port 161 is opened on the SCOM 2012 installation. This is the reason why SCOM 2012 isn't able to act as a SNMP trap receiver out of the box unless the device discovery is successful done!
[root@<hostname> ~]# nmap -sU -p U:161 <FQDN-SCOM> Starting Nmap 4.11 ( http://www.insecure.org/nmap/ ) at 2014-06-20 12:28 CEST Interesting ports on <FQDN-SCOM> (<SCOM-IP>): PORT STATE SERVICE 161/udp open|filtered snmp MAC Address: 00:50:56:A3:xx:yy (VMWare) Nmap finished: 1 IP address (1 host up) scanned in 0.325 seconds [root@<hostname> ~]# [root@<hostname> ~]# nmap -sU -p U:162 <FQDN-SCOM> Starting Nmap 4.11 ( http://www.insecure.org/nmap/ ) at 2014-06-20 12:28 CEST Interesting ports on <FQDN-SCOM> (<SCOM-IP>): PORT STATE SERVICE 162/udp closed snmptrap MAC Address: 00:50:56:A3:xx:yy (VMWare) Nmap finished: 1 IP address (1 host up) scanned in 0.118 seconds [root@<hostname> ~]#
But now let us configure the MS Windows Server 2012 and the SCOM 2012 R2 update rollup 2 in several separate steps.
Click on Server Manager -> Dashboard -> Add roles and features
Read and click next
Select role-based of feature-based installation
Select a server from the server pool
Just click next
Check SNMP service (should be installed)
Click on Server Manager -> Tools -> Services
Click on SNMP Service -> Properties
Set SNMP community
Select accept SNMP packets from any host
SNMP Service startup type has to be automatic and Service status has to be running
SNMP Trap startup type has to be disabled and Service status has to be stopped
Click on Authoring -> Management Pack Objects -> Rules -> Create a new rule...
Select SNMP Trap (Event) rule type and click on New... near Management pack
Create the Management pack
Enter some knowledge (or not) and click on Create
Back on rule type selection click on Next
Select a rule target first
Select Node as target
Select rule name
Configure the trap OIDs to collect
Enable the scope view
Select the target "Node"
Review the settings of the created rule of type node
"Collect all SNMP traps" properties
SNMP Trap Event trap provider properties
Click on Monitoring -> New -> Folder...
Review the settings of the created rule of type node
Create an Event View
Edit the properties of the Event View
Click on Administration -> Run As Configuration -> Create Run As Account...
Specify general account properties
Select a distribution security option
Prompt to create a profile
Click on Administration -> Run As Configuration -> Profiles -> Create Run As Profile...
Specify general profile properties
Add Run As Account
Warning about account distribution
Click on Administration -> Network Management -> Discovery Wizard...
Select a management or gateway server
Check discovery rules
Set general properties
Set general properties
Select a discovery type
Specify the default Run As accounts for discovery
Schedule the network discovery
Confirm the settings
Wait for completion
And voila there is a new network SNMP version 1 device discovered on port 161!
A nmap scan again shows that port 162 is now listening on the SCOM 2012 system.
[root@<hostname> ~]# nmap -sU -p U:162 <FQDN-SCOM> Starting Nmap 4.11 ( http://www.insecure.org/nmap/ ) at 2014-06-20 13:31 CEST Interesting ports on <FQDN-SCOM> (<SCOM-IP>): PORT STATE SERVICE 162/udp open|filtered snmptrap MAC Address: 00:50:56:A3:xx:yy (VMWare) Nmap finished: 1 IP address (1 host up) scanned in 0.321 seconds [root@<hostname> ~]#
Now it's time to disable (stop) the SNMP daemon on the Linux server and try to send some SNMP traps. There is a command line utility which can be used for sending SNMP traps. Don't confound things, this is not snmptrapd!
[root@<hostname> ~]# service snmpd stop Stopping snmpd: [ OK ] [root@<hostname> ~]#
[root@<hostname> ~]# snmptrap -v 1 -c public <FQDN-SCOM> .188.8.131.52.4.1.8072.2.3.1 "" 6 17 "" .184.108.40.206.220.127.116.11.0 s "Just there" [root@<hostname> ~]#
[root@<hostname> ~]# snmptrap -v 1 -c public <FQDN-SCOM> NET-SNMP-EXAMPLES-MIB::netSnmpExampleNotification "" 6 17 "" SNMPv2-MIB::sysLocation.0 s "Just here" [root@<hostname> ~]#
shows following result in SCOM. Please note that SCOM doesn't translate the OIDs to the appropriate entity names by itself!
For the interested reader now a small collection of SNMP v2c and v1 traps and how SCOM detected them.
Trap version 2(c) SNMP trap
Use of snmptrap cli:
[root@<hostname> ~]# snmptrap -v 2c -c public <FQDN-SCOM> 0 NET-SNMP-EXAMPLES-MIB::netSnmpExampleHeartbeatNotification netSnmpExampleHeartbeatRate i 60 [root@<hostname> ~]#
[root@<hostname> ~]# snmptrap -v 2c -c public <FQDN-SCOM> 0 .18.104.22.168.4.1.8072.2.3.0.1 .22.214.171.124.4.1.8072.2.3.2.1.0 i 60 [root@<hostname> ~]#
Display into SCOM 2012:
Trap version 1 SNMP trap
Use of snmptrap cli:
[root@<hostname> ~]# snmptrap -v 1 -c public <FQDN-SCOM> NET-SNMP-EXAMPLES-MIB::netSnmpExampleNotifications "" 6 1 "" netSnmpExampleHeartbeatRate i 60 [root@<hostname> ~]#
[root@<hostname> ~]# snmptrap -v 1 -c public <FQDN-SCOM> .126.96.36.199.4.1.8072.2.3 "" 6 1 "" .188.8.131.52.4.1.8072.2.3.2.1.0 i 60 [root@<hostname> ~]#
Display into SCOM 2012:
Now we see a bunch on non-speaking object identifiers. There is a tool called snmptranslate into the net-snmp package which translates the OID into the entity name and vice versa:
|ObjectIdentifier||snmptranslate -m ALL||Syntax||Value||snmptranslate -m ALL|
If you use the option -m ALL snmptranslate will scan the MIB directory, RHEL default is /usr/share/snmp/mib, parse all MIBs and build the entity names out of the OIDs.
I've bundled the needed MIBs into a download file [vmware-mibs.tar.gz;82 kByte], which you can request here. More information about the latest MIBs are available from VMware: http://kb.vmware.com/kb/1013445.
This was a short trip into the world of object identifiers and entity names. Let us now switch back to real world. We are now able to discover a network device, which sends SNMP traps to SCOM 2012. But what about network devices which are unable to answer SNMP requests but able to send SNMP traps? Is it possible to monitor them with System Center Operations Manager?
The answer is: yes, but not the smart way. First you have to deploy a server like the RedHat Enterprise Linux I've mentioned above and configure it identical (hostname and ip address and SNMP version) to your network device. If this Linux server was successful discovered by SCOM just remove it and substitute it with the "real" SNMP trap sending network device. This should work pretty good! But beware the circumstance that SCOM 2012 just shows the object identifier and is not able to import MIBs! This is a severe shortage.
I guess some guys know that the snmptrap daemon is able to forward SNMP traps. Now, maybe that it is possible to forward SNMP traps from a MS Windows server, let's say VMware vCenter Server, using this RHEL server? Well I've extend my scenario to verify this:
Linux A sends a snmptrap to Linux B. Linux B is configured as a forwarder for SNMP traps and sends the trap immediatly back to Linux A which is a snmptrap receiver, too. I do this without including SCOM to verify that the setup is full functional. We can monitor the behavior of the two systems by observing the content of /var/log/messages on both servers.
send trap from Linux A to Linux B:
[root@<hostnameA> ~]# snmptrap -v 1 -c public <hostnameB> .184.108.40.206 localhost 6 17 "" .220.127.116.11 s "Just a test" [root@<hostnameA> ~]#
Logfile Linux B:
Jun 18 14:46:04 <hostnameB> snmptrapd: 2014-06-18 14:46:04 localhost.localdomain [127.0.0.1] (via UDP: [<hostnameA-IP>]:38153) TRAP, SNMP v1, community public SNMPv2-SMI::internet Enterprise Specific Trap (17) Uptime: 265 days, 5:59:58.77 SNMPv2-SMI::internet = STRING: "Just a test"
Logfile Linux A:
Jun 18 14:46:04 <hostnameA> snmptrapd: 2014-06-18 14:46:04 localhost.localdomain [127.0.0.1] (via UDP: [<hostnameB-IP>]:37360) TRAP, SNMP v1, community public SNMPv2-SMI::internet Enterprise Specific Trap (17) Uptime: 265 days, 5:59:58.77 SNMPv2-SMI::internet = STRING: "Just a test"
Hey, that looks good.
/var/log/messages from Logfile Linux A:
Jun 18 14:47:00 <hostnameA> snmptrapd: 2014-06-18 14:47:00 <vCenter-FQDN> [<vCenter-IP>] (via UDP: [<vCenter-IP>]:56912) TRAP, SNMP v1, community public VMWARE-PRODUCTS-MIB::vmwVC Enterprise Specific Trap (VMWARE-VC-EVENT-MIB::vpxdDiagnostic) Uptime: 34 days, 21:34:31.53
Isn't it cool? It will work pretty good from the Linux point of view. Let's transfer the scenario using SCOM 2012 as SNMP trap receiver, which means that there is an additional trap receiver to configure:
As you might guess, the result is: nothing! I can't see any new trap into SCOM 2012 by using the last scenario, just the old ones sending direct from Linux B. Oh, what a mess!
These are the configuration changes on Linux B:
authCommunity net,log public forward default <hostnameA-IP> public forward default <SCOM-IP> public traphandle VMWARE-VC-EVENT-MIB::vpxdDiagnostic /root/scom/bin/traps vpxdiag traphandle VMWARE-PRODUCTS-MIB::vmwVC /root/scom/bin/traps vcenter traphandle .18.104.22.168.4.1.6876.4.3 /root/scom/bin/traps vcenteroid traphandle .22.214.171.124.4.1.68126.96.36.199.202 /root/scom/bin/traps vpxdiagoid disableAuthorization yes
You see the forward rules, they are self-explaining. The next lines are trap handlers, I will tell a bit later about them. The last line (disableAuthorization yes) is needed if you use net-snmp v5.3 or greater.
The trap handlers are able to execute a script if a specific trap is received. Because the forwarding doesn't work, what about triggering a new trap with the received values? Let's see what happens if we create the script.
#!/bin/bash BASE='/usr/bin/snmptrap -v 1 -c public <SCOM-FQDN> .188.8.131.52.4.1.6876.4.3 "" 6' VPXDIAG='202 ""' VPXDALARMINFO='203 "" ' vars= while read oid val do OBJECT="$VPXDALARMINFO" if [ "$oid" = "<vCenter-FQDN>" ] then continue fi if [ "$oid" = "<UNKNOWN>" ] then continue fi if [ "$oid" = "UDP:" ] then continue fi if [ "$oid" = "SNMP-COMMUNITY-MIB::snmpTrapAddress.0" ] then continue fi if [ "$oid" = "SNMP-COMMUNITY-MIB::snmpTrapCommunity.0" ] then continue fi if [ "$oid" = "DISMAN-EVENT-MIB::sysUpTimeInstance" ] then continue fi if [ "$oid" = "SNMPv2-MIB::snmpTrapOID.0" ] then continue fi if [ "$oid" = "SNMPv2-MIB::snmpTrapEnterprise.0" ] then continue fi if [ "$oid" = "VMWARE-VC-EVENT-MIB::vmwVpxdTargetObjType.0" ] then if [ "$val" = "unknown" ] ; then val=1 elif [ "$val" = "host" ] ; then val=2 elif [ "$val" = "vm" ] ; then val=3 elif [ "$val" = "other" ] ; then val=4 fi fi if [ "$val" = "VMWARE-VC-EVENT-MIB::vpxdDiagnostic" ] then OBJECT="$VPXDIAG" break fi if [ "$vars" = "" ] then if [ "$oid" = "VMWARE-VC-EVENT-MIB::vmwVpxdTargetObjType.0" ] then vars="$oid i \"$val\"" else vars="$oid s \"$val\"" fi else if [ "$oid" = "VMWARE-VC-EVENT-MIB::vmwVpxdTargetObjType.0" ] then vars="$vars $oid i \"$val\"" else vars="$vars $oid s \"$val\"" fi fi done echo "trap: $1 # $BASE $OBJECT $vars" >> /root/scom/bin/snmptraphandle.txt clean_vars=$(echo $vars | sed -e 's/Ä/Ae/g' -e 's/Ö/Oe/g' -e 's/Ü/Ue/g' -e 's/ä/ae/g' -e 's/ö/oe/g' -e 's/ü/ue/g' -e 's/ß/ss/g') eval "$BASE" "$OBJECT" "$clean_vars"
... . /etc/init.d/functions OPTIONS="-m ALL -Ls 3 -p /var/run/snmptrapd.pid" #OPTIONS="-Lsd -p /var/run/snmptrapd.pid" if [ -e /etc/sysconfig/snmptrapd.options ]; then ...
[root@<hostnameA> ~]# service snmptrapd restart Stopping snmptrapd: [ OK ] Starting snmptrapd: [ OK ] [root@<hostnameA> ~]#
To receive the vpxd diagnostic trap from vCenter every minute follow the kb article about how to change the default sending period of the diagnostic SNMP trap on vCenter: http://kb.vmware.com/kb/2020271. If you don't know how to send SNMP traps from vCenter there is a kb article, too: http://kb.vmware.com/kb/1006438.
Yeah, we've got it! It works as expected. Now we can receive SNMP traps send by a vCenter Server running Microsoft Windows using a Linux server as a kind of trap rewriter.
Now let me write some words about the configuration. You may have recognized in snmptrapd.conf the duplicate lines, e.g.
traphandle VMWARE-VC-EVENT-MIB::vpxdDiagnostic /root/scom/bin/traps vpxdiag
traphandle .184.108.40.206.4.1.68220.127.116.11.202 /root/scom/bin/traps vpxdiagoid
In your final configuration you could delete one of the lines. It doesn't matter if you use OIDs or entity names. In my opinion it's better to use the entity names, because they are more "human readable" instead of the OIDs. Additionally you can delete the forward lines, they are obsolete.
I was inspired from the handler script example from the net-snmp team for writing the traps shell script. The script is just a kind of filter which throws away unwanted information. It converts the ASCII string of the vmwVpxdTargetObjType to integer again (the MIB files provide information about the needed data type) and at least it coverts the german umlauts (some guys say "funny or dotted characters") to the valid ASCII range. If you don't do that SCOM interpretes this as the binary representation of octet string as you can see in the following picture:
Just by replacing the umlaut ä to "ae" the octet string get readable again:
Now my dear reader is it up to you to build a free piece of software which displays the vCenter SNMP traps in a meaningful way into SCOM 2012. I've provided you with all information and tips I've currently discovered.
You can download this page as pdf file [2000 kB].
Am Dattenbach 9-11
|Phone: not available|
|frank [dot] ickstadt [at] removethis gmail [dot] com||Fax: currently out of order|