Skip to content

Change service dashboard's SLI Detail rows to start from y=200

Gregorius Marco requested to merge mg-fix-service-dashboard-layout into master

To test:

$ ./test-dashboard.sh sidekiq/main.dashboard.jsonnet
Installed https://dashboards.gitlab.net/dashboard/snapshot/qEZypRZBzwZp9e62iMFCPcYFuEeclkAT - sidekiq: Overview

This MR fixes a visual bug found in GitLab.com Sidekiq service dashboard.

In Sidekiq service overview dashboard, the y grid from panels inside Service Level Indicators row overflows to the rows from SLI Detail: * rows.

$ cat generated/sidekiq/sidekiq-main.json | jq '.panels[] | .title +  - y:  + (.gridPos.y|tostring)'
...
"shard_urgent_authorized_projects SLI Apdex - y: 121"
" - y: 128"
"shard_urgent_authorized_projects SLI Error Ratio - y: 121"
" - y: 128"
"shard_urgent_authorized_projects SLI RPS - Requests per Second - y: 121"
"Details - y: 121"
"shard_urgent_cpu_bound SLI Apdex - y: 131"
" - y: 138"
"shard_urgent_cpu_bound SLI Error Ratio - y: 131"
" - y: 138"
"shard_urgent_cpu_bound SLI RPS - Requests per Second - y: 131"
"Details - y: 131"
"shard_urgent_other SLI Apdex - y: 141"
" - y: 148"
"shard_urgent_other SLI Error Ratio - y: 141"
" - y: 148"
"shard_urgent_other SLI RPS - Requests per Second - y: 141"
"Details - y: 141"
"🔬 SLI Detail: email_receiver - y: 100"
"🔬 SLI Detail: external_dependency - y: 110"
"🔬 SLI Detail: global_search_indexing - y: 120"
"🔬 SLI Detail: shard_catchall - y: 130"
...

Here, the shard_urgent_other SLI RPS - Requests per Second ends at y=141 whereas SLI Detail starts at y=100

Visually, it looks like this. Notice the intermingled row and panels. image

This is because Sidekiq has too many SLIs (1 SLI component per shard) and 1 SLI takes 10 units.

This fix doesn't really matter though once gitlab-com/gl-infra/scalability#2317 (closed) replaces the per-shard SLIs with just 2 SLIs (execution and queueing)

Edited by Gregorius Marco

Merge request reports