Hi everyone,
I'd like to share an experience and ask for advice regarding monitoring an Aruba CX cluster in VSX-Gateway mode. In our setup, the two cluster members act as the network core and are located in separate data center rooms.
As you know, in VSX mode, both members share the virtual IP address (the "active-gateway"). However, there is no specific OID to poll for the traffic of this virtual IP.
To work around this, I discovered that the OIDs ipIfStatsInOctets (1.3.6.1.2.1.4.31.3.1.5.1.{#SNMPINDEX}) and ipIfStatsOutOctets (1.3.6.1.2.1.4.31.3.1.32.1.{#SNMPINDEX}) can be used to retrieve traffic data from the local VLAN interfaces on each member.
My solution was to create two calculated items on one of the cluster member hosts (which acts as my monitoring proxy). These items sum the latest values from these two OIDs retrieved from both cluster members. This sum gives me the total bandwidth for the entire cluster.
However, the implementation hasn't been straightforward. In the latest data, I've observed that the values polled from one or both members occasionally drop to 0 for several update intervals. This is unusual, as network devices don't typically behave this way. This results in graphs with frequent "holes" wherever a value of 0 is recorded.
To mitigate this, I added a "discard unchanged" preprocessing step to my items. The preprocessing pipeline is as follows:
This has smoothed the graphs somewhat, but the bandwidth lines still lack continuity.
Has anyone else encountered this situation with Aruba VSX or similar clustering technologies? If so, how did you work around it?
Thank you for your feedback!
I'd like to share an experience and ask for advice regarding monitoring an Aruba CX cluster in VSX-Gateway mode. In our setup, the two cluster members act as the network core and are located in separate data center rooms.
As you know, in VSX mode, both members share the virtual IP address (the "active-gateway"). However, there is no specific OID to poll for the traffic of this virtual IP.
To work around this, I discovered that the OIDs ipIfStatsInOctets (1.3.6.1.2.1.4.31.3.1.5.1.{#SNMPINDEX}) and ipIfStatsOutOctets (1.3.6.1.2.1.4.31.3.1.32.1.{#SNMPINDEX}) can be used to retrieve traffic data from the local VLAN interfaces on each member.
My solution was to create two calculated items on one of the cluster member hosts (which acts as my monitoring proxy). These items sum the latest values from these two OIDs retrieved from both cluster members. This sum gives me the total bandwidth for the entire cluster.
However, the implementation hasn't been straightforward. In the latest data, I've observed that the values polled from one or both members occasionally drop to 0 for several update intervals. This is unusual, as network devices don't typically behave this way. This results in graphs with frequent "holes" wherever a value of 0 is recorded.
To mitigate this, I added a "discard unchanged" preprocessing step to my items. The preprocessing pipeline is as follows:
- Change per second
- Discard unchanged
- Custom multiplier (8, to convert Octets to bits)
This has smoothed the graphs somewhat, but the bandwidth lines still lack continuity.
Has anyone else encountered this situation with Aruba VSX or similar clustering technologies? If so, how did you work around it?
Thank you for your feedback!