Feature #47072: mgr/dashboard: Usability Improvements
mgr/dashboard: landing page
Current landing page, despite the recent refresh, still exhibits some issues:
- Effective visualization
- There are 17 widgets loosely grouped around 3 themes (Status, Capacity & Performance).
- Status cards are mostly text, static and with no visual hints. They also occupy a third of the landing page. This could be condensed and undesired/changing situations could be visually highlighted.
- Missing data:
- In light of the ability to enable compression and the future deduplication feature the dashboard should be able to show a savings/storage efficiency ‘card’
- Pool status and usage.
- Data becoming less relevant/obsolete:
- PG’s per OSD are not that relevant with the autoscaler, and should be removed (as long as we have an out-of-the-box alert in place)
- Scrubbing (are operators really monitoring here whether scrubbing happens or not? this should perhaps be an alert, rather than a fixed card 99% in inactive state)/
- The alerts shown should also have links to the relevant component (OSD DOWN -> OSDs page for example).
This would be chance for another rethinking on the landing page (v3) in order to make it more intuitive, more evenly distributed when it comes to information displayed and its relevancy.
Current Landing Page:
The design doc is available here#.
#4 Updated by Paul Cuzner over 2 years ago
There are a few things worth highlighting in the mockup to avoid confusion
- by placing all the components in a single place, it's visually easier to understand and locate info
- by using up/down/maintenance indicator we improve the status information we present to the user ( a host that is down for example doesn't show at all in the dashboard)
- by using consistent state icons for up/down/maintenance/drain we provide a more intuitive lifecycle approach to the components we show.
- any problems (daemon/hosts down) would be shown as a hyperlink to the relevant section in the UI
- we already show small line graphs in pool and osd pages, so these widgets would just be based on that existing code
- the time window for the chart would be fixed - the intent is to show the last couple of minutes only - any more than that, the admin can use the grafana charts
- putting health front and centre presents an easy focal point for the new user
- depicting health based on PG state is sufficient for most users of a UI (CLI users and devs always want more!)
- the alerts listed don't show the other concept of linking from the landing page. Today we link Pools to the pools page for example, which is fine, but I think for alerting we should be doing the same thing - alert osd down...link to osd page to provide the new user with a more intuitive experience.
- visualizing capacity as a bar (or something!) that includes the nearfull and full ratios provides immediate indication of usage - answering the question, how far away am I from capacity shortages..
- Compression has been around for a while and IMO, we should be advertising the storage efficiency features of ceph, not hiding them
- deduplication is slated to be new in Pacific so introducing a storage efficiency tile, gets ahead of this new feature and provides a home for any stats that it can give us.