DX NetOps

Hi,

Please let me know if a device goes down when polled,Operators are alerted via the event list or other mechanism to take an immediate action to a downed device.When the next polling cycle hits, the device responds.   Was it a false alarm?Can we rapid poll the device and if it was a false alarm it can be cleared in a few seconds, rather than minutes.  Operators could then know that a device failure may have been a false alarm.

For example, if polling is set for 2 min.Then rapid polling should be done for every 10 seconds on downed device.

Is there any concept to do this in Spectrum 10.2.

The operators can be notified by different ways:

• When a new alarm is generated, the operator can be notified by a sound.

       General Preferences - CA Spectrum - 10.2 to 10.2.3 - CA Technologies Documentation 

       New Message Sound
       Specifies whether there is a sound indicator when you receive a new message from OneClick administrators.

• When a new alarm is generated, the operator can be notified by email.

       How to configuring Spectrum Alarm Notifier to send - CA Knowledge 

Please take a look at Tech Tip:

Tech Tip: Spectrum Device Communication Timeout Intelligence Concepts 

Thanks for the response Silvio.

I have small doubt,do  DCM(Device Communication manager) acts on the devices that are polled first and stated as down or it will act on all the devices.

For example, I have 100 devices in which one device is generated alarm as down.Do DCM act on all the 100 devices or it tries to communicate with one device that is down.

Thanks,

Sheethal

Hi Sheethal,

Suppose you have this default setting:

• Poll Interval (sec): 300
• DCM Timeout (ms): 3000
• DCM Retry Count: 3

CA Spectrum polls each device every 5 minutes (300 seconds).

In each poll (SNMP get-request), CA Spectrum will wait for 3 seconds (3000 ms) for a response. If the device did not respond in 3 seconds, CA Spectrum will poll the device again, up to 3 times.

If there is no response for SNMP get-request after 3 attempts, CA Spectrum will ping the device. If the device replies to ping, but not to SNMP, CA Spectrum asserts the "MANAGEMENT AGENT LOST" alarm.

But if the device does not respond to ping either, CA Spectrum asserts the "DEVICE HAS STOPPED RESPONDING TO POLLS" alarm.

If the connectivity of your network model is complete, CA Spectrum Fault Isolation will polls the problematic device's neighbor to identify the most probable fault device.

Fault Isolation - CA Spectrum - 10.2 to 10.2.3 - CA Technologies Documentation 

Thanks,

Silvio

Additionally, Spectrum staggers it's polling workload among the devices.  That means if all the devices are polled @ 300 seconds (5 min), device A might be polled at 12:00, 12:05, 12:10 but device B might be polled at 12:01, 12:06, 12:11.  If both device A and B are connected to each other, if A is polled successfully @ 12:00 but B fails @ 12:01, A will be polled again before 12:05 because Spectrum is trying to determine neighbor connectivity with B.

The minimum interval for polling devices in spectrum is 30 seconds. You cannot poll devices faster than 30 seconds. The link that Silvio provided describes how the timeouts in device communications are handled.

You can go lower than 30 seconds, if you like but it's not recommended.  There's obviously increased load on Spectrum when you do that but you also run into situations where the SNMP counters on the device may not update that frequently or has resource problems keeping up with the requests.  It was many years ago but I've seen Spectrum crush Catalyst switches with polling requests if you drop it too low (lower than 10 seconds). 

At least in OneClick, if you try to set a value of lower than 30 seconds, you'll get a pop-up like this one:

it's true, I did not try to modify it using CLI or rest.

That's fairly recent, at least in the larger timeline of Spectrum.  It's probably a good thing to protect users from accidentally crushing their network. 

Just to let you know, Spectrum is already doing that by default. Have a look here: Poll Devices That Are Down - CA Spectrum - 10.2 to 10.2.3 - CA Technologies Documentation 

 

When contact with a device has been lost, CA Spectrum uses two methods to continue polling the device for a status change. First, CA Spectrum pings the device once every 60 seconds. Second, CA Spectrum sends SNMP requests every third polling interval by default. For example, if the device polling interval is set to 60 seconds, CA Spectrum polls the down device once every 180 seconds.

To change the default interval at which CA Spectrum polls a down device, insert the following syntax into the <$SPECROOT>/SS/.vnmrc file:

down_device_poll_interval_multiplier=<user_defined_multiplier>

For example, if the <user_defined_multiplier>=2 and the device polling interval is 60 seconds, then CA Spectrum polls the down device once every 120 seconds (2*60=120).

Hope this helps!

We have a concept in spectrum, but I don't want to wait till next polling cycle I want to confirm the alarm generated is false alarm or device is really down before operator takes an action.

Are you using Spectrum integrated with UIM? For UIM models only icmp will work for availability .SNMP will not come into .

No my spectrum is not integrated with UIM.