Ceph » mgr » Dashboard

Description of problem¶

Just tried out the dashboard with the landing page revamp. Looks promising, but there are a number of issues that make it problematic at the moment.
Obviously this is a subjective view, but

• Status card - When there is an event, I can see the icon change, but there is no information about the event i.e healthcheck name. This seems to be a regression
• Used Capacity is wrong - there is already a tracker for that, which is great.
• Under the Used capacity title it just shows used - recommend adding a value for total or even a % full value
• Cluster utilization card title should probably be Cluster Utilization (i.e. capitalize)
• tooltip on IOPS is wrong (shows OPS and IPS) - should it be read and write?
• the actual IOPS shown for reads and writes is incorrect - I ran a rate limited workload fixed to 20 reads a 2 writes IOPS - the chart shows something like 120 and 18 - I think the query is incorrect? Is it using osd_op_*_ bytes?
• tooltip on Latency is wrong - says read/write - but I believe it's showing the commit and apply latency averages across all OSDs
• tooltips on the charts are confined to the chart divs so they get cut-off obscuring or hiding information (firefox 111) - if this is a problem could the presented to the left or right of the datapoint instead of above and below?
• client throughput tooltip on the chart should show the Unit (eg KiB/s)
• recovery throughput should also show the unit
• IOPS and client throughput values (on the left) showing 0, but the chart is plotting non-zero values?
• Use capacity showing the wrong unit (I had a cluster with 1.5 TB, and it was showing a y-axis of 400TB (there was 326MB used at this point)
• The IOPS and throughput chart tooltip behavior only shows a single value at a time, but the latency chart shows both. Having both values revealed within the tooltip is more useful.
• Hovering over anywhere in the cluster utilization area shows a tooltip stating Cluster Utilization - even over the charts, which IMO is wrong.

Also I think the title "Latency" is misleading - as an admin I'd assume this is client latency - but the since the plot is apply/commit I'd recommend renaming to "OSD Apply/Commit Latency"

Environment¶

• ceph version string: ceph version 18.0.0-3146-gcec15a59 (cec15a5992a4d51d286e5cb510d667af3b27c07a) reef (dev)
• Platform (OS/distro/release): CentOS9stream
• Cluster details (nodes, monitors, OSDs): All-on-one deployment from main
• Did it happen on a stable environment or after a migration/upgrade?: Testing
• Browser used (e.g.: Version 86.0.4240.198 (Official Build) (64-bit)): Firefox 111 (Linux/Fedora)