Resources
The Resources section of the menu holds three pages: Usage, what everything takes right now; Statistics, the same over time; and Storage, where the disk has gone.
Usage
The Usage page puts the whole server on one screen. Three charts split the CPU, the memory and the disk between the instances, the core services, PostgreSQL and the host itself, with what is still free. Below them, Where it goes lists the host first - its reserve, the core services and PostgreSQL container by container, the Docker images, the archive - and then every instance with its CPU, its memory against its ceiling and its disk. Open an instance for its own breakdown: its container, its databases, filestore, backups, add-ons and logs. The breakdown is read when you open it, so a server with many instances opens quickly.
CPU and memory are read live. Disk comes from the hourly measurement; Refresh reads the CPU and memory again at once and, for an administrator, measures the disk again too. An operator or viewer scoped to some instances sees those instances only, and not the host.
The same table explains one instance's memory on its Configuration tab, under Where the memory goes.
Statistics
The Statistics page shows CPU, memory and disk for the host and for each managed container, as sparklines over a rolling window. Samples are collected continuously and kept in a ring buffer, so the page answers "what has this server been doing" rather than only "what is it doing now".
A reading with a known capacity turns to Warning at 80% of it and to Critical at 90%. Both are set under Configuration, as Usage warning (%) and Usage critical (%), and they are one pair for the whole console: the Storage page and the dashboard colour a filesystem by them, and the resource and low-disk alerts are raised by them.
This is the honest check on tuning: worker counts sized for a peak that never happens cost memory every hour of every day.
Storage
The Storage page breaks down where space has gone: the instance data trees, backups, archives, Docker images, volumes and build cache.
Docker is usually the surprise. Per-instance images accumulate as instances are rebuilt, and the build cache grows quietly. Removing images an instance no longer uses is one of the built-in maintenance operations.
pelmo$ ohm instance image list --instance fanespelmo$ ohm instance image remove --instance fanesSizes
pelmo$ ohm instance size --instance fanespelmo$ ohm instance database size --instance fanesBackups are the other half
The Storage page counts your backup directory only if it is on this server. Local backups on separate storage, and everything at an FTP, S3 or Azure target, are outside its view.
The copies kept are set per target for this reason: a local target keeping a week and an off-site target keeping a year is a normal arrangement, and the local disk should be sized for the week. See Copies kept.