Host map
Every host as a tile, grouped and coloured, so the outlier is visible before you read a number.
The map #
Each host is one tile, sized by the group it belongs to and shaded by a metric. The point is to see the machine that stands out without reading a table.
Shading #
The fill is a sequential ramp over one metric, so a cluster running hot reads as a block of colour rather than a set of numbers.
- CPU %
- Memory %
- Disk %
- Load
Health is shown separately, as a glyph on the tile and a labelled badge on hover. A busy host is not necessarily an unhealthy one, so the two are never mixed.
Grouping #
Group tiles by any one of:
- Cluster
- Region
- Service
- Provider
- Env
Grouping by cluster is the default and answers 'which cluster is hot'. Grouping by service answers 'which service's fleet is hot'.
Reading it #
- A block of dark tiles is a cluster under pressure.
- One dark tile among pale neighbours is the machine to open.
- A group badge glyph marks hosts reporting errors in the window.
- Hovering a tile gives the value it is shaded on.
Finding a host #
Type into the filter to narrow by host name, service or cluster. It searches the tile labels, so it finds the host even when it sits inside a large group.
From the map to the machine #
Opening a tile takes you to that host's detail, where the load figures and the containers running on it are waiting.
The map reads metrics, the list reads identity
A host can be listed without appearing on the map, because the map needs a metric to shade it by. Check Hosts first when a host is missing from the map but expected on it.
Start grouped by cluster, then switch to service once you know which cluster it is. The first view narrows the fleet, the second tells you whose it is.
Onward #
The tile is the shortlist; Hosts is the detail. For what the machine is running, Containers.