Zabbix network volume issue

[KitCB]

I have been using Zabbix successfully for a couple of years to monitor devices on remote networks. It is a simple installation that is currently monitoring 123 devices distributed over 18 different remote networks. The server connects to each of these networks via IPSEC links. The network where the server is located uses a slow 1mbs dedicated link for IPSEC traffic. The only traffic that passes over this link to each of the 18 networks is either Zabbix traffic or administrator remote access maintenance traffic. (i.e. on demand ssh traffic or rdp traffic for troubleshooting etc.) Each device is being monitored by the unmodified OS template (i.e. Linux/Windows/BSD) for the device.

Recently, I have been experiencing significant delays when working across this connection performing maintenance tasks. I have determined that the data between the Zabbix server and the devices is consuming all of the bandwidth for the connection. With all other traffic removed from the connections, I am seeing that the Zabbix server is sending and receiving over 2.5 gigabytes of data in each direction daily.

I have searched through the forums, but I can't seem to find anything that:

1 - identifies bandwidth requirements between server and client. (i.e. how much data is transferred between the server and each client when performing standard checks? Is my reported bandwidth to be expected for the number of devices etc.?)

2 - I can increase or decrease the number of pollers, trappers and timers but I can't find any parameter in the zabbix_server.conf file to adjust the timing of the pollers. In my usage case I am really just checking that each device is alive and well. While real-time notifications are desirable, I'd be happy to trade delayed checks with a reduction in bandwidth usage.

Any suggestions? (I realize that the easy answer is just add more bandwidth).

[batchenr]

I have been using Zabbix successfully for a couple of years to monitor devices on remote networks. It is a simple installation that is currently monitoring 123 devices distributed over 18 different remote networks. The server connects to each of these networks via IPSEC links. The network where the server is located uses a slow 1mbs dedicated link for IPSEC traffic. The only traffic that passes over this link to each of the 18 networks is either Zabbix traffic or administrator remote access maintenance traffic. (i.e. on demand ssh traffic or rdp traffic for troubleshooting etc.) Each device is being monitored by the unmodified OS template (i.e. Linux/Windows/BSD) for the device.

Recently, I have been experiencing significant delays when working across this connection performing maintenance tasks. I have determined that the data between the Zabbix server and the devices is consuming all of the bandwidth for the connection. With all other traffic removed from the connections, I am seeing that the Zabbix server is sending and receiving over 2.5 gigabytes of data in each direction daily.

I have searched through the forums, but I can't seem to find anything that:

1 - identifies bandwidth requirements between server and client. (i.e. how much data is transferred between the server and each client when performing standard checks? Is my reported bandwidth to be expected for the number of devices etc.?)

2 - I can increase or decrease the number of pollers, trappers and timers but I can't find any parameter in the zabbix_server.conf file to adjust the timing of the pollers. In my usage case I am really just checking that each device is alive and well. While real-time notifications are desirable, I'd be happy to trade delayed checks with a reduction in bandwidth usage.

Any suggestions? (I realize that the easy answer is just add more bandwidth).

im no expert but what i think is maybe you should go throw your items
ans check which items you can increase their "update interval" to make checks less frequently and maybe it will reduce the bandwidth that zabbix takes you.

[SBO]

Hi,

In the requirements section of Zabbix, you can find the size of an item, to estimate the size of your DB : https://www.zabbix.com/documentation...#database_size

I suppose you can take that to calculate your bandwidth usage too.

For your bandwidth issue, you said you are using the default templates.
As you mainly want to know if your nodes are alive, you should remove these templates and simply put the ICMP ping one.

It will avoid to query date for cpu, disk, memory, etc.., and will just do a ping every X seconds (30s by default IIRC, but you can of course adjust it).

[KitCB]

I was hoping that there would be a way to globally change the interval, but I think realistically, changing the interval in the templates is going to be my best solution.

[KitCB]

Using the calculation from the requirements document:

The dashboard reports that my required, new values per second is 59.54

So, 59.54 * 90 bytes *86400 seconds per day results in 462,983,040 bytes per day or approximately 462MB.

I am sure there is also some overhead involved but this seems quite a bit less than the 5GB of data I am seeing transferred on a daily basis.