System status
Everything is running
Component health for the Isobar platform, refreshed every 30 seconds from the same probes our on-call uses.
99.920%
30-day availability
92
monitored components
2
incidents this quarter
12min
median time to mitigate
Components
Component healthall systems operational
Ingest (ECMWF)operational
60 days ago99.976% uptimetoday
Ingest (NCEP)operational
60 days ago99.964% uptimetoday
Rechunking pipelineoperational
60 days ago99.938% uptimetoday
Subsetting APIoperational
60 days ago99.898% uptimetoday
Zarr direct accessoperational
60 days ago99.960% uptimetoday
Job queueoperational
60 days ago99.917% uptimetoday
Incident history
Every incident since the platform launched, including the ones customers never noticed.
Delayed cache purges in the EU region
resolved27 Jun 2026 · impact: no data loss
14:52 UTC
Root cause identified: a faulty line card on one of two upstream ports. Traffic was drained from the affected device.
14:21 UTC
We are investigating elevated p99 latency reported by monitoring in a single facility.
15:40 UTC
The device was replaced and traffic re-balanced. Metrics are back to baseline; we are keeping the incident open for another hour to confirm.
Scheduled maintenance — border router firmware
resolved01 Jun 2026 · impact: none
02:00 UTC
Maintenance window opened. Requests were served by the standby control plane throughout.
03:12 UTC
Upgrade completed with no customer-visible impact.
About this page
Synthetic probes run from nine independent networks on three continents. A component counts as down when at least three probes on different networks fail two consecutive checks.
Anything that degrades a customer-visible SLO: availability, error rate, or the latency budget for the affected component.
Yes — the status API returns the same data as JSON, and every component exposes a Prometheus-compatible metrics endpoint.