How this works: architecture, safety system, and the chaos engineering game
Network Topology
VPN (active)
eBGP K8s
ClusterMesh
Upstream
DMZ
VPN Tunnels
NL ↔ GR
budget
40ms
NL ↔ NO
budget
20ms
NL ↔ CH
budget
25ms
NL ↔ NO-DMZ01
budget
—
NL ↔ NO-DMZ02
budget
—
NL ↔ GR
freedom
40ms
NL ↔ NO
freedom
20ms
NL ↔ CH
freedom
25ms
NL ↔ NO-DMZ01
freedom
—
NL ↔ NO-DMZ02
freedom
—
GR ↔ NO
inalan
60ms
GR ↔ CH
inalan
55ms
GR ↔ NO-DMZ01
inalan
—
GR ↔ NO-DMZ02
inalan
—
NO ↔ CH
vps
22ms
NO ↔ NO-DMZ01
vps
—
NO ↔ NO-DMZ02
vps
—
CH ↔ NO-DMZ01
vps
—
CH ↔ NO-DMZ02
vps
—
NO-DMZ01 ↔ NO-DMZ02
pair
—
NL ↔ TX
budget
120ms
NL ↔ TX
freedom
120ms
GR ↔ TX
inalan
155ms
CH ↔ TX
vps
145ms
NO ↔ TX
vps
140ms
TX ↔ NO-DMZ01
vps
—
TX ↔ NO-DMZ02
vps
—
27/27
VTI Tunnels
27 active
85/87
BGP Established
3
Failover Layers
READY
ClusterMesh
6 global svc
78ms
Avg Latency
p99 217ms
227.1s
MTTR
38 events / 7d
GR
secondary
NO
transit
Tunnels
18/18 up
Uptime
100.00%
CH
transit
Tunnels
7/7 up
Uptime
100.00%
TX
transit
Tunnels
7/7 up
Uptime
100.00%
Cross-Site Latency (ms)
| NL | GR | NO | CH | TX | |
|---|---|---|---|---|---|
| NL | — | 40.0 | 20.0 | 25.0 | 120.0 |
| GR | 40.0 | — | 60.0 | 55.0 | 155.0 |
| NO | 20.0 | 60.0 | — | 22.5 | 140.0 |
| CH | 25.0 | 55.0 | 22.5 | — | 145.0 |
| TX | 120.0 | 155.0 | 140.0 | 145.0 | — |
Failover Layers
1
BFD sub-second (~900ms)
2
BGP hold-time 30s (aggressive VTI timers 10/30)
3
ISP SLA-track default route (~6-10s)
Last: WAN failover · Jun 9, 06:55 UTC · 38 events/24h