You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Recently we noticed high traffic rates in iptables counters on master servers. After investigation we found that the main part of traffic comes from consul agents on port 8300 (RPC).
This creates a big problem in environments with limited bandwidth or with a large number of nodes.
A little context. We have a cluster that consists of more than 100 nodes (nomad + consul) that are geographically located in different parts of the world and communicate via a VPN channel.
We noticed that on smaller clusters (6 nodes, but many services on each node) - the situation is repeated and found that about 1-1.5 GB of traffic comes to the master per day on port 8300 from 1 agent.
Thus, this creates problems with large distributed clusters with 100+ nodes (it turns out that we receive more than 100 GB of traffic every day only from the consul to the master server) because of which we have already started receiving letters about exceeding the limits (20 TB / month)
Can you explain what could be the cause of such a volume of traffic or is this normal behavior?
I will provide the configuration of 1 master and 1 agent (some values are edited for security purposes)
I would also like to note that there are no messages in the Consul logs except information (we also turned on debug, but there was nothing interesting there either). Consul quite often synchronizes the traefik service (more often than the others), namely the service, not the check, but I doubt that this service is the cause of such high traffic consumption.
Reproduction Steps
Steps to reproduce this issue, eg:
Create a cluster with 100+ client nodes and 3 or more server nodes in different subnets. Connection via vpn is not necessary. it is enough for agents and masters to see each other directly
Set up separate counters for clients and masters network on port 8300. (also you could setup other counters for rest consul ports)
Let it for 1h and see high net usage (if *24 to get value for 1d)
We tried to reduce consumption by adding the following parameters to the consul configuration (they can be ignored, the problem is reproducible without them)
Overview of the Issue
Recently we noticed high traffic rates in iptables counters on master servers. After investigation we found that the main part of traffic comes from consul agents on port 8300 (RPC).
This creates a big problem in environments with limited bandwidth or with a large number of nodes.
A little context. We have a cluster that consists of more than 100 nodes (nomad + consul) that are geographically located in different parts of the world and communicate via a VPN channel.
We noticed that on smaller clusters (6 nodes, but many services on each node) - the situation is repeated and found that about 1-1.5 GB of traffic comes to the master per day on port 8300 from 1 agent.
Thus, this creates problems with large distributed clusters with 100+ nodes (it turns out that we receive more than 100 GB of traffic every day only from the consul to the master server) because of which we have already started receiving letters about exceeding the limits (20 TB / month)
Can you explain what could be the cause of such a volume of traffic or is this normal behavior?
I will provide the configuration of 1 master and 1 agent (some values are edited for security purposes)
I would also like to note that there are no messages in the Consul logs except information (we also turned on debug, but there was nothing interesting there either). Consul quite often synchronizes the traefik service (more often than the others), namely the service, not the check, but I doubt that this service is the cause of such high traffic consumption.
Reproduction Steps
Steps to reproduce this issue, eg:
We tried to reduce consumption by adding the following parameters to the consul configuration (they can be ignored, the problem is reproducible without them)
Consul info for both Client and Server
Client info
Server info
Operating system and Environment details
Run on ubuntu 24.04 LTS
The text was updated successfully, but these errors were encountered: