# Build resources and cache > Limit the CPU, memory and disk builds use on a server, give builds more time, and free up disk space. For agents: [llms.txt](https://ployz.dev/llms.txt) indexes the Ployz docs as Markdown. Builds on your servers share CPU, memory and disk with your services. A big build can slow your app down, so you can cap what builds use on each server. Start in the dashboard: lower a server's **Builds at once**, give builds their own server, or move them to GitHub Actions. See [Where builds run](https://ployz.dev/docs/builds/where-builds-run). For finer limits, use the settings below. They can't be set from the dashboard yet: you set them on the server, over SSH as root. Builds on GitHub Actions don't use them. ## Cap CPU, memory and cache Create `/root/.ployz/build.yaml` on the server. Every line is optional: ```yaml # Each build may use at most 2 CPU cores cpu_cores: 2 # ...and at most 4 GiB of memory, in bytes. A build that needs more fails. memory_bytes: 4294967296 # Keep the build cache near 20 GiB, in bytes cache_bytes: 21474836480 # Keep at least 10 GiB of disk free, in bytes min_free_bytes: 10737418240 ``` The next build on that server uses it; nothing needs a restart. A mistake in the file fails every build on the server with `invalid host build.yaml: …` or a message that names the field, so watch the next build after you change it. - **The limits are per build.** With **Builds at once** set to 2, builds on the server can use twice as much. - **Cache sizes are targets.** A big build can briefly go over them. - **Without `cache_bytes` or `min_free_bytes`**, Ployz keeps a fifth of the server's Docker disk free. ## Give builds more time A build on your servers fails after 30 minutes. To allow more, up to 86400 seconds (a day), add a line to `/etc/default/ployz` on the server, then restart Ployz. Above 86400, Ployz won't start. To fail a build that waits too long for a free slot, also add `PLOYZ_BUILD_QUEUE_TIMEOUT_SECONDS`, with the same limit. Without it, a build waits as long as it takes. Restarting drops any builds waiting on the server, so do it when it isn't building: ```sh # Builds may run for 1 hour (3600 seconds) echo 'PLOYZ_BUILD_ACTIVE_TIMEOUT_SECONDS=3600' >> /etc/default/ployz systemctl restart ployz ``` ## Free up disk space The build cache makes builds faster. To free its space and start from an empty cache, run this on the server as root when it isn't building: ```sh ployz server build-cache-clear ``` It keeps your images and everything else in Docker. After each deploy, Ployz cleans up old images it built, but never images it pulled for [Docker image services](https://ployz.dev/docs/deploy/docker-image). Remove those with `docker image prune` when you need the space.