OpenWrt Infrastructure Meeting August 2026

Date: August 26 2026

Present: Baptiste, Ansuel, Ted, Adrian, Paul

Next meeting: 23th September

Ted: Upgrade is a bit stuck because of disk space needed for the SQL migration.

Ted will resize the VM to have enough space for the upgrade.

Baptiste did some recent actions to improve the stability:

• added 20 GB of nginx cache in front of cgit
• lowered CPU priority for cgit, so that git clone fails less often
• increased number of cgit workers
• disable tarball generation

The cgit web interface is still not very reliable, but git clone fails much less often now.

TODO: add monitoring for cache size

It would be interesting to understand where git traffic comes from (user-agent / IP)

It would also be interesting to separate metrics for cgit and for git operations

For now, we only monitor git-03 and robimarko-01 (setup done by hand).

Work in progress to deploy it everywhere with Ansible:

• Wireguard tunnels
• install node-exporter on all nodes
• configure node-exporter to listen on wireguard interface only
• add all nodes to victoriametrics config to scrape metrics

Second step: nginx monitoring

Third step: alerting rules (e.g. to get an email if disk space is low on a build worker)

Fourth step: service-specific metrics (sysupgrade server, discourse, anubis on wiki...)

Fifth step: expose simple public dashboards (e.g. download statistics)

TODO:

• Ted will install the Discourse plugin to generate Prometheus metrics, and Baptiste/Paul will integrate it in Grafana
• Baptiste and Paul will do a work session to finish the first step (Ansible setup)

Still needs some love.