Ask one interface, get answers from another ...

6 messages Options
Embed this post
Permalink
Klaus Steden

Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink

Hi there,

I've got a device here that's does something that confuses OpenNMS. It has
two interfaces on the same network. One is a management interface, the other
is a data interface. It is used as an ASI/IP bridge (we use it to do WAN
transmission of SDI video).

I've noticed when I let OpenNMS autodiscover it, it will add both nodes,
then incorporate the management node as an interface under the data node's
entry. I did a bit of head scratching, and when I ran tcpdump while doing an
snmpwalk against the managment interface, I noticed that while the SNMP
query is sent to the management interface, it's the data interface that
replies. I don't know how common this is, but it's producing confusing
information in OpenNMS.

Additionally, while it presents the Net-SNMP OIDs with information about
packets, throughput, etc., and this is visible with snmpwalk, none of this
information ever gets graphed in OpenNMS.

To add insult to injury, the sysObjectID.0 is a very helpful '.0.0', instead
of the vendor-specific OID. It's not under support and we're not likely to
renew it, so I probably have to live with this mistake, but my questions
are:

1. how do I monitor the machine, so that OpenNMS doesn't try to query the
management interface for any services but still tracks it via ICMP for node
up/down status, i.e. which interface do I want to query, and which one do I
ignore, or should I just let OpenNMS figure that out?

2. does the bogus sysObjectID.0 value cause a problem to OpenNMS in general?
Is there anything I can do to address the issue, if it is a problem?

3. what would cause OpenNMS to fail to collect more data than just basic
response time from a node that's exporting the IF, IP, and SNMPv2-SMI MIBs?

thanks,
Klaus


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss
Michael Seibold

Antw: Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink
Hi Klaus,

> It has two interfaces on the same network

Can you set a host route on this device pointing to your management interface on this machine to force it to send all answers to your management server through the same interface?

> 1. how do I monitor the machine, so that OpenNMS doesn't try to query the
> management interface for any services but still tracks it via ICMP for node
> up/down status, i.e. which interface do I want to query, and which one do I
> ignore, or should I just let OpenNMS figure that out?

I think service discovery should be configured in capsd-configuration.xml  and service (icmp...) polling in poller-configuration.xml. Discovering the services won't harm normally, so just disable polling for them.


...

> 3. what would cause OpenNMS to fail to collect more data than just basic
> response time from a node that's exporting the IF, IP, and SNMPv2-SMI MIBs?

It has to find a package in datacollection-config.xml where it can match the sysoid from your device! No match -> no data.


- Michael


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss
jcat

Re: Antw: Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink
Some javascript/style in this post has been disabled (why?)
You could just bind the snmp daemon to one IP address for completion, then add the route to the ONMS server that specifies the particular interface you want to reply from.



Cheers,
Just

On Thu, 2009-11-05 at 14:27 +0100, Michael Seibold wrote:
Hi Klaus,

> It has two interfaces on the same network

Can you set a host route on this device pointing to your management interface on this machine to force it to send all answers to your management server through the same interface?

> 1. how do I monitor the machine, so that OpenNMS doesn't try to query the
> management interface for any services but still tracks it via ICMP for node
> up/down status, i.e. which interface do I want to query, and which one do I
> ignore, or should I just let OpenNMS figure that out?

I think service discovery should be configured in capsd-configuration.xml  and service (icmp...) polling in poller-configuration.xml. Discovering the services won't harm normally, so just disable polling for them.


...

> 3. what would cause OpenNMS to fail to collect more data than just basic
> response time from a node that's exporting the IF, IP, and SNMPv2-SMI MIBs?

It has to find a package in datacollection-config.xml where it can match the sysoid from your device! No match -> no data.


- Michael


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss


--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss
Klaus Steden

Re: Antw: Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink
In reply to this post by Michael Seibold


On 11/5/09 5:27 AM, "Michael Seibold" <[hidden email]> etched on
stone tablets:

> Hi Klaus,
>
>> It has two interfaces on the same network
>
> Can you set a host route on this device pointing to your management interface
> on this machine to force it to send all answers to your management server
> through the same interface?
>
I don't believe so, no ... the network stack is implemented in VxWorks, and
is somewhat brittle in nature. The UI it presents only allows setting a
default gateway, although oddly enough the way this information is entered,
it gives the appearance that the default gateway applies only to the
management interface, not the data interface.

>
>> 1. how do I monitor the machine, so that OpenNMS doesn't try to query the
>> management interface for any services but still tracks it via ICMP for node
>> up/down status, i.e. which interface do I want to query, and which one do I
>> ignore, or should I just let OpenNMS figure that out?
>
> I think service discovery should be configured in capsd-configuration.xml  and
> service (icmp...) polling in poller-configuration.xml. Discovering the
> services won't harm normally, so just disable polling for them.
>
What seems to happen is OpenNMS eventually concludes that the management
interface is actually associated with the node for the data interface, and
treats it as a part of the data interface node, rather than as a separate
node. I'm not sure if this is by design or a side-effect, but if it's by
design, I think it's a good design, so cheers on that. :-)
>
> ...
>
>> 3. what would cause OpenNMS to fail to collect more data than just basic
>> response time from a node that's exporting the IF, IP, and SNMPv2-SMI MIBs?
>
> It has to find a package in datacollection-config.xml where it can match the
> sysoid from your device! No match -> no data.
>
Okay, that makes sense. I have a couple of disparate systems that also use a
sysObjectID of .0.0 -- there's an Apple XRAID enclosure that's similarly
brain-dead. I added this to the systems section of datacollection-config.xml
and re-scanned, but then the node stopped showing the 'Update SNMP' link
altogether so I'm still confused:

-- cut --
<systemDef name="Broken SNMP Implementations">
  <sysoid>.0.0</sysoid>
  <collect>
    <includeGroup>mib2-host-resources-system</includeGroup>
    <includeGroup>mib2-host-resources-memory</includeGroup>
    <includeGroup>mib2-host-resources-storage</includeGroup>
    <includeGroup>net-snmp-disk</includeGroup>
    <includeGroup>net-snmp-proc</includeGroup>
    <includeGroup>ucd-loadavg</includeGroup>
    <includeGroup>ucd-memory</includeGroup>
    <includeGroup>ucd-sysstat</includeGroup>
    <includeGroup>ucd-sysstat-raw</includeGroup>
    <includeGroup>ucd-sysstat-raw-more</includeGroup>
  </collect>
</systemDef>
-- cut --

Additionally, I now see this message in the collectd.log (sorry for the
crappy linewrap):

-- cut --
2009-11-05 23:49:19,374 WARN  [CollectdScheduler-50 Pool-fiber2]
CollectableService: Unable to initialize 10.99.42.93 for SNMP collection,
reason: Interface 10.99.42.93 is not the primary SNMP interface for nodeid
161
2009-11-05 23:49:19,447 WARN  [CollectdScheduler-50 Pool-fiber4]
CollectableService: Unable to initialize 10.99.42.93 for SNMP collection,
reason: Interface 10.99.42.93 is not the primary SNMP interface for nodeid
161
2009-11-05 23:50:30,966 WARN  [btpool0-9] IfLabel: Interface
(nodeId/ipAddr=161/10.99.42.93) has no ifName and no ifDescr...setting to
label to 'no_ifLabel'.
-- cut --

10.99.42.93 is the IP address of the data interface, i.e., the one that
answers SNMP queries -- should I change the way the systemDef is matched or
should using '<sysoid>.0.0</sysoid>' match these broken implementations that
don't properly assign an enterprise OID number to the sysObjectID? Is more
information needed in datacollection-config.xml to get OpenNMS to gather
SNMP data from these nodes?

thanks,
Klaus


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss
jcat

Re: Antw: Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink
Some javascript/style in this post has been disabled (why?)
That is by design, when ONMS queries the interface tables from the SNMP daemon it finds that both IPs are part of the same node and joins them.


Cheers,
Just

On Fri, 2009-11-06 at 00:31 -0800, Klaus Steden wrote:

What seems to happen is OpenNMS eventually concludes that the management
interface is actually associated with the node for the data interface, and
treats it as a part of the data interface node, rather than as a separate
node. I'm not sure if this is by design or a side-effect, but if it's by
design, I think it's a good design, so cheers on that. :-)


--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss
Klaus Steden

Re: Antw: Ask one interface, get answers from another ...

Reply Threaded More More options
Print post
Permalink

... any idea why the system now fails to collect SNMP, or what kind of
definition to use (<sysoid> vs. <sysoidMask>) for systems that have
broken/stupid default sysObjectID values?

Klaus

On 11/6/09 1:22 AM, "jcat" <[hidden email]> etched on stone tablets:

> That is by design, when ONMS queries the interface tables from the SNMP
> daemon it finds that both IPs are part of the same node and joins them.
>
>
> Cheers,
> Just
>
> On Fri, 2009-11-06 at 00:31 -0800, Klaus Steden wrote:
>
>
>> What seems to happen is OpenNMS eventually concludes that the management
>> interface is actually associated with the node for the data interface, and
>> treats it as a part of the data interface node, rather than as a separate
>> node. I'm not sure if this is by design or a side-effect, but if it's by
>> design, I think it's a good design, so cheers on that. :-)
>
>


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Please read the OpenNMS Mailing List FAQ:
http://www.opennms.org/index.php/Mailing_List_FAQ

opennms-discuss mailing list

To *unsubscribe* or change your subscription options, see the bottom of this page:
https://lists.sourceforge.net/lists/listinfo/opennms-discuss