[IOS XR] SNMP Monitor Graph not stable

[anas.abdullkarim]

hello , i have user PRTG for Monitoring my router ASR9001 

i have added snmp interface sensors and CQOS sensors 

the problem is sometime PRTG have a stable traffic Graph and sometime Zigzag Graph 

we my router in peak-load traffic , that is happened 

[smilstea]

How often are you polling the router for IFMIB?

Do you have ifmib caching enabled?

1. Without ifmib caching it takes longer to get results back as we have to poll the interfaces one at a time to get real data.

2. With ifmib caching SNMP will poll the cached value not the actual value which is quicker, the downside is that the values in the cache are only updated every 30s.

Sam

[anas.abdullkarim]

thanks for helping , 
i have not idea about ifmib caching

how to enable it in cisco 9001 (xr os ) 

[smilstea]

It would show under 'show run snmp-server'.

If it is or isn't there we can try toggling it to see what results you get in your collector after that.

Sam

[smilstea]

https://www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k_r4-0/system_management/command/reference/yr40asr9k_chapter12.html#wp375831232

[anas.abdullkarim]

thanks , u enabled it , but same problem to now , the graph is like Zigzag 

[mromanini]

Anyone have the solution to this problem? 

I have the exact same issue. We have ASR9k  6.2 showing values like zizag. For example i have an interface police rate at 4 gbps against a upstream provider and some times y get a lecture like 3.5 gbps and the next one of 4.5 gbps. Its seems like it is triying compensate the previos reading with a higher value. I have tested it with ifmib caching and without but the problem persist.

In the begining I thought it could be a problem of zabbix but it is not. I made a script that polls every 5 minutes and I get the exact sames results that zabbix does.

I have a multivendor environment and this behavior its only seen with ios xr.

Any sugestions? thank in advance

[smilstea]

what is the interface load-interval set to?

Sam

[mromanini]

Hi, thanks for the answer.

The load interval is 5 minutes.

[smilstea]

Can you try changing the load-interval to the minimum? I believe that is 30s. Also what OID are you polling?

[mromanini]

I'm polling IF-MIB::ifHCInOctets.

I did change the load-interval to 30 secs and now its seems to be working find.

Can you explain why was that? as per my understanding the curve should have been smoother than polling every 30 sec.

Thank you very much

[smilstea]

When you change the load-interval it changes how often the statistics are updated, which is why you saw the ping ponging before.

https://www.cisco.com/c/en/us/td/docs/routers/asr9000/software/system_monitoring/command/reference/b-sysmon-cr-asr9k/b-sysmon-cr-asr9k_chapter_0111.html

Sam