Hosts tab: show CPU, disk and memory per host #8

New issue

Open

opened 2026-10-20 07:55:59 +00:00 by rosa · 0 comments

rosa commented

2026-10-20 07:55:59 +00:00

Member

Requested by Felix 2026-10-20. The Hosts tab should show per-host resource usage, not just liveness.

What to show, per host

CPU: load average and/or utilisation %.
Memory: used / total, and swap if present.
Disk: used / total for the partition that actually matters (the data/workspace volume, not /).
Plus freshness: when the sample was taken, so a stale number is visibly stale rather than silently wrong.

Collection. The runtime already talks to each host over SSH for provisioning and updates (hostprovision.py), so a periodic lightweight sample (e.g. cat /proc/loadavg, free -b, df -B1 <path>) fits the existing transport. Avoid adding an agent where SSH already works.

Design points that matter

Staleness is the main risk. A host that stops reporting must render as unknown, never as its last known value — that is how a dead host looks healthy.
Keep the sample on the runtime side and push it into the host document, so the web tier stays a reader.
Threshold warnings (disk >85%) would have caught the quota incident directly.

Requested by Felix 2026-10-20. The Hosts tab should show per-host resource usage, not just liveness. **What to show, per host** - CPU: load average and/or utilisation %. - Memory: used / total, and swap if present. - Disk: used / total for the partition that actually matters (the data/workspace volume, not `/`). - Plus freshness: when the sample was taken, so a stale number is visibly stale rather than silently wrong. **Collection.** The runtime already talks to each host over SSH for provisioning and updates (`hostprovision.py`), so a periodic lightweight sample (e.g. `cat /proc/loadavg`, `free -b`, `df -B1 <path>`) fits the existing transport. Avoid adding an agent where SSH already works. **Design points that matter** - **Staleness is the main risk.** A host that stops reporting must render as unknown, never as its last known value — that is how a dead host looks healthy. - Keep the sample on the runtime side and push it into the host document, so the web tier stays a reader. - Threshold warnings (disk >85%) would have caught the quota incident directly.

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: NSTERN/TreeGent#8

No description provided.

Rows
Columns