Operating IPv6: monitoring, measurement and SLAs
Monitor IPv6 and IPv4 separately for reachability, latency and traffic share. Define IPv6 SLAs and alerting. Report IPv6 adoption progress with flow data (RFC 7011).
Take the quiz · pass mark 70% · 9 questions
Explanation
Once IPv6 is deployed it must be operated as a first-class protocol — which means measuring it on its own, not hidden behind IPv4.
Monitor each family separately
The single most important operational rule: probe IPv4 and IPv6 independently. A dashboard that just says "the service is up" tells you nothing about IPv6, because the probe (like any client) may be using IPv4. Run per-family checks — reachability, latency, loss and throughput for each of the two protocols — using tools that let you force the family (a blackbox exporter configured for IPv6, curl -6/curl -4, per-family synthetic transactions). This is the operational form of the L6.08 testing rule: Happy Eyeballs hides IPv6 faults, so you must look at IPv6 directly.
Traffic share and flows (RFC 7011)
To know how much of your traffic is actually IPv6, export flow records. IPFIX (RFC 7011, collector port 4739) carries per-flow data including the IPv6 source/destination information elements, so a collector can report the IPv6 vs IPv4 byte/packet share, top talkers per family, and trends over time. This is both an operational view (capacity, anomalies) and the data behind the adoption metric the enterprise/ISP program (L7.03) tracks.
IPv6 SLAs and alerting
Write SLAs that name IPv6 explicitly: reachability and latency targets per family, so an IPv6-only degradation breaches the SLA even while IPv4 is fine. Alerting must fire on per-protocol divergence (IPv6 down or slow while IPv4 is up) — exactly the condition a combined check misses. Translate availability targets into concrete budgets: for a 30-day month, a 99% target allows 432 minutes of downtime, a 99.9% target about 43 minutes — compute the budget and alert before you burn it.
Reporting adoption
Report progress to stakeholders with a small, stable set of numbers: percentage of IPv6 traffic (from IPFIX), percentage of sites/services dual-stacked, per-family availability against SLA, and the IPv6 incident count/MTTR. A monthly trend of IPv6 traffic share is the single most convincing adoption metric.
ictNetworks scenario
ictNetworks runs a Prometheus blackbox exporter probing every service over IPv4 and IPv6 separately, with Grafana dashboards and alerts that fire on per-protocol divergence. An IPFIX collector (port 4739, RFC 7011) reports the IPv6 traffic share and top talkers per family, feeding both capacity planning and the monthly adoption report. Its SLAs state per-family reachability and latency with downtime budgets (99% = 432 min/month), and IPv6 incidents are tracked with their own MTTR.
Drawing diagram…
Worked examples
Example 1: The IPFIX collector port
Flow records are exported to an IPFIX collector on port 4739 (RFC 7011); the IPv6 address information elements let the collector split traffic by family.
Example 2: Families to probe
Honest IPv6 operations means probing 2 address families separately for every service, because a combined "is it up" check can pass on IPv4 while IPv6 is broken.
Example 3: The downtime budget
A 99% monthly availability SLA over a 30-day month permits 30 * 24 * 60 * 0.01 = 432 minutes of downtime. Converting the percentage to a minute budget lets you alert on error-budget burn before the SLA is breached.
Lab
Goal: Build a dual-stack monitoring dashboard with per-protocol alerting.
You need: Prometheus, the blackbox exporter, Grafana, an IPFIX-capable router/collector, and a few dual-stack services. (All vary by version — check the docs.)
- Configure the blackbox exporter to probe each service over IPv6 and IPv4 separately (two modules/targets per service).
- Build a Grafana dashboard showing per-family reachability and latency side by side.
- Export IPFIX (RFC 7011) to a collector and add a panel for the IPv6 vs IPv4 traffic share and top talkers.
- Write alerts that fire on per-protocol divergence (IPv6 down/slow while IPv4 is up) and on SLA error-budget burn (using the computed minute budget).
- Break IPv6 for one service (e.g. drop its AAAA or block ICMPv6) and confirm the IPv6-specific alert fires while IPv4 stays green.
What to submit: the exporter/Prometheus/alert configuration, the dashboard, the IPFIX traffic-share panel, and evidence that the per-protocol alert fired on the injected IPv6 fault.
Summary
- Operate IPv6 as a first-class protocol: probe IPv4 and IPv6 separately for reachability, latency and throughput.
- Export IPFIX (RFC 7011, port 4739) to measure the IPv6 traffic share and feed the adoption metric.
- Write per-family SLAs and alert on per-protocol divergence and error-budget burn (99%/30 days = 432 min).
- Report adoption with IPv6 traffic share, dual-stack coverage, per-family availability and incident MTTR.
- This completes Level 7; next is Level 8 (advanced topics and capstone).
Quiz imported · 9 questions · pass mark 70%. Lab questions are not auto-marked.