Test and Monitoring
Testing for connectivity
It's important to validate that your instance is configured correctly, and that users can connect to them. You can use our connection tester to run through a series of tests and confirm that it's seeing the expected results.
Performance tuning
To achieve high performance, some tuning is necessary.
If you are deploying with our Kubernetes Helm chart, these settings will be applied automatically.
Host networking
If running in Docker, ensure that you are using host networking. NAT and bridge mode introduce another layer of translation that could be limiting to performance.
Kernel parameters
Several kernel parameters would need to be tuned via sysctl. By default, Linux comes with conservative settings on buffer/queue sizes that would lead to packet loss under load.
These values can be set in /etc/sysctl.conf (and will take effect upon next boot), or applied immediately via sysctl.
# Increase number of file handles
fs.file-max = 2097152
# Increase number of incoming connections
net.core.somaxconn = 65535
# Increase size of the receive queue
net.core.netdev_max_backlog = 65535
# Increase maximum amount of option buffer to 25M
net.core.optmem_max = 25165824
# Increase socket read and write buffer maximum to 25M
net.core.rmem_max = 25165824
net.core.wmem_max = 25165824
# Increase socket read and write buffer maximum to 1M
net.core.rmem_default = 1048576
net.core.wmem_default = 1048576
To apply via sysctl, use:
$ sudo sysctl -w net.core.rmem_max=25165824
net.core.rmem_max = 25165824
Prometheus
When configured, LiveKit exposes Prometheus stats via /metrics endpoint. We export metrics related to rooms, participants, and packet transmissions.
To view the list of metrics we export, see roomstatsreporter.go