Hello Everyone,
I am sure folks have figured out the right solution to this, but my 'google-fu' is failing me and I cannot seem to find the answer.
We have a number KVM servers in our environment, running Centos 7. The virtual NICs created by libvirt all report a speed of 10Mb/s, and this is what gets reported back via SNMP to zambix for all of these NICs after the discovery phase. The problem is that these NICs are really 1G and so we get lots of alarms when the VMs on the host get a little busy and the throughput exceeds what Zabbix thinks the speed of the interface is.
According to Redhat: "vnet interfaces are tap devices. The 10Mbps is just a cosmetic figure in a tap device's virtual firmware. This speed has no bearing on the speed of a guest's virtual device whatsoever."
So...there is really no hope of Redhat changing this as they obviously don't see it as a problem.
What do other folks do to handle this problem:
1) Go into each discovered vnet NIC in Zabbox and hardcode a speed in the trigger calculations?
2) Change the discovery process so that vnet NICs are always 1Gbps?
3) Disable the triggers for vnet NICs and just let the triggers for the non-virtual NICs be my guide?
If the answer is #1 or #2...can someone point me to the relevant portion of the docs for me to make the necessary changes?
Thanks,
Emmett
I am sure folks have figured out the right solution to this, but my 'google-fu' is failing me and I cannot seem to find the answer.
We have a number KVM servers in our environment, running Centos 7. The virtual NICs created by libvirt all report a speed of 10Mb/s, and this is what gets reported back via SNMP to zambix for all of these NICs after the discovery phase. The problem is that these NICs are really 1G and so we get lots of alarms when the VMs on the host get a little busy and the throughput exceeds what Zabbix thinks the speed of the interface is.
According to Redhat: "vnet interfaces are tap devices. The 10Mbps is just a cosmetic figure in a tap device's virtual firmware. This speed has no bearing on the speed of a guest's virtual device whatsoever."
So...there is really no hope of Redhat changing this as they obviously don't see it as a problem.
What do other folks do to handle this problem:
1) Go into each discovered vnet NIC in Zabbox and hardcode a speed in the trigger calculations?
2) Change the discovery process so that vnet NICs are always 1Gbps?
3) Disable the triggers for vnet NICs and just let the triggers for the non-virtual NICs be my guide?
If the answer is #1 or #2...can someone point me to the relevant portion of the docs for me to make the necessary changes?
Thanks,
Emmett
Comment