Skip to content

Ops dashboard ​

The sample Bastion screen: status bar, edge-region latency map, services table, deploy timeline, a resolved incident, a log tail, and an ops-agent suggestion with Approve and Dismiss. Figures match the approved layout. They are not a live cluster.

Change the region, pause the tail, refresh metrics, open an incident or the deploy list (Escape closes it), then approve or dismiss the suggestion. Approving sets worker to 4/4 on this page only.

SAMPLEFictitious telemetry. Approve, dismiss, refresh and region changes stay on this page.

BASTION

ALL SYSTEMS NOMINAL

REGION
UPTIME99.982%
Edge regionsLatency map
Sample edge latencyAshburn 12 milliseconds, Chicago 18 milliseconds, Amsterdam 15 milliseconds, Frankfurt 15 milliseconds, São Paulo 45 milliseconds, Singapore 67 milliseconds, Hong Kong 71 milliseconds, Sydney 78 millisecondsiad12msord18msams15msfra15mssao45mssin67mshkg71mssyd78ms
< 20ms20–50ms> 50ms
ServicesOverview
Sample service status, CPU, memory and replicas
ServiceStatusCPUMemoryReplicas
api RUNNING
18%
41%
3/3
worker RUNNING *
62%
67%
2/3
db RUNNING
24%
58%
2/2
cache RUNNING
11%
32%
2/2
* High CPU utilization detected
DeploysTimeline
Newest on the left. Amber means review, not only color.
IncidentsRecent

INC-2025-0417-03 RESOLVED

High latency on ap-southeast-1

Root cause: Elevated packet loss on upstream provider.

Resolved by rerouting traffic via alternate peering.

Detected Apr 17 09:12 · Resolved Apr 17 09:47

Sample incident. No paging.
LogsTail(bastion-ops-agent)
  1. [INFO]Ops agent health check passed
  2. [INFO]All regions nominal
  3. [INFO]Worker queue depth: 128
  4. [WARN]High CPU on worker (62%)
  5. [INFO]Scaling recommendation generated
  6. [INFO]Suggested action: scale worker to 4 replicas
  7. [INFO]Awaiting operator confirmation…
Tail streaming sample lines
Bastion ops agent

Ops Agent sugiere escalar worker a 4 réplicas

Razones:

  • CPU utilization sostenido > 60% en los últimos 10m
  • Cola de trabajo creciendo (depth: 128)

Confianza del agente: 82%

INC-2025-0417-03

RESOLVED

High latency on ap-southeast-1.

Root cause: elevated packet loss on the upstream provider. Resolved by rerouting via alternate peering.

Detected Apr 17 09:12 · Resolved Apr 17 09:47. Sample record only.

Sample deploys

  • 12:42 · a7f3c2d · v2.17.4
  • 11:03 · b81d9e4 · v2.17.3
  • 09:31 · c9b2a1f · v2.17.2
  • 08:05 · d2e4f6a · v2.17.1 · review

Local list. No release was shipped.

What is on the screen ​

RegionRole
Status barWordmark, ALL SYSTEMS NOMINAL, region select, uptime and sparkline
Latency mapDotted field, eight sample cities, legend for <20ms, 20–50ms, and >50ms
ServicesStatus, CPU, memory, replicas. Worker starts at 62% CPU and 2/3 replicas
DeploysFour hashes. d2e4f6a is marked Review, not color alone
IncidentINC-2025-0417-03, resolved sample
Log tailbastion-ops-agent, with pause
Ops agentSpanish recommendation, 82% confidence, Approve and Dismiss

Responsive ​

At desktop widths the panels sit in two columns. Around a tablet width, and whenever the frame is under 860px, they stack. The service table scrolls inside its panel on a narrow phone. Status controls wrap instead of forcing the page sideways.