Table of Contents

Pollers: General Overview, Deployment Strategy, Build Out & Setup

Morgan Wynnes Updated by Morgan Wynnes

What is a Poller:

A Poller is a device that usually sits within your network to monitor and collect network information on deployed devices and bring that data back to Sonar. Additionally a Poller will be able to determine which deployed devices on accounts are connected to which equipment on network sites.

One Poller can monitor about one to two thousand devices if enough resources are given. However, a Poller can possibly perform worse depending on your network build out and what kind of equipment you have deployed. The requirements for a poller is 4 Gigs of Ram and a dual core CPU with Ubuntu 16.04, and while having a more powerful PC help with polling, it is recommended to have 2 or 3 pollers if your network is larger than 3 or more thousand devices.

Deployment Strategy:

The Poller should be hosted within your network rather than outside (Digital Ocean, AWS or Azure). While you can host your Poller on the cloud, it is recommended to be as close to your deployed devices as possible. There are 2 reasons why this is recommended, first being that you will get a quicker response from the devices you are polling and second it is more secure than hosting it on the cloud.

Additionally if you have a large network that spans a large area, it is recommended to put the pollers as close as possible to the devices you are polling. This ensures that you have accurate information.

Build out:

To create a Poller in Sonar, go to setting monitoring and click on Pollers. Then click on the create button and create a poller. In here name the poller to something recognizable (a good rule of thumb is to name the poller on where it is located or what subnets it is polling) and choose the subnets you want to monitor. In here you can also set the priority of that subnet. The lower the priority the more accurate information you are getting for that subnet. By pressing the + button you can add more subnets to be polled by that poller.

Once a poller has been created, you will be taken back to the poller screen with all your pollers. In here you can see the API key which we will need this key later.

From here we can install the Poller on a virtual machine. The virtual machine should have 4 Gigs of RAM, at least 2 cores, and 20 Gigs of Hard Drive Space.

Additionally we need to set how often we will poll info on accounts and tower equipment. This information is set in Poller Settings.

  • Account Polling: Determines how often deployed equipment on accounts are polled. This should not be set to anything lower than 5 minutes unless you have a very small network.
  • Infrastructure polling: Determines how often deployed equipment on tower sites are polled. generally we want to poll this info every 2 to 5 minutes.

Instructions on how Installing the Poller on a Virtual Machine can be found on our GitHub

Troubleshooting steps:

Poller has not returned any data to Sonar in the past 5 minutes:

  • Check the API key and the link to the Sonar instance to make sure it is correct.
  • Poller is polling too many devices or too frequently. Uncheck some of the subnets to see if it makes any difference and if so deploy additional pollers. Additionally you might have to increase your polling settings for accounts.
  • Poller is running out of space. Check if debug mode is On by running nano /opt/poller/.env and if so turn it off. Check the hard drive space by running df -h .

Device in Sonar shown in a down state:

  • Check to see if both ICMP and SNMP are down. If just SNMP then the device is not down and it is showing a false positive.
  • Run SNMP walk on poller to make sure the device is able to respond to SNMP requests. If it does not make sure SNMP is enable on the device and the community string matches what is defined on the template. Additionally make sure you have not overwritten the SNMP version or the community string on the device.
  • If only certain devices are down (especially Ubiquity) increase the SNMP_timeout to 1.5 to 2 seconds and increase the SNMP retries to 1.

How did we do?

Building a Monitoring Template

Building Alerting Rotations