Ingesting DNS telemetry
Configure your consumer using Connecting a Kafka consumer and the shared SIEM ingestion guide. Use the DNS feed schemas for field types, action catalogs, and complete payload examples.
Map operations and platform logs
Parse each Kafka value as a JSON object. The operations and platform feeds provide these fields where applicable:
| Field | SIEM use |
|---|---|
feed | Feed and parser selection when several feeds share one SIEM input |
schema_version | Payload-contract version; currently 1 for every DNS feed |
timestamp | Source event time |
customer_id | Customer scope |
server | Source host |
zone | DNS asset or zone; absent for some repository-level operations |
actor, actor_type | Executing or observing component |
action | Event type for rules and dashboards |
log_level | Severity |
message | Analyst display text |
event_id | Record identifier present in all four feeds; use it for replay deduplication |
production_sha | Approved configuration revision in operations records |
approved_by | Approver identities in applicable operations records |
serial | Zone version when supplied |
Set timestamp as event time and retain its original nine fractional digits
in the raw record. Store ingestion time separately. See
Time provenance for the
atomic-clock reference and DNS alignment assurance. Use action and
structured fields for rules rather than parsing message.
Any feed can deliver duplicate records after a reconnect or replay. Every copy
retains the same event_id; deduplicate by event_id alone. Log feeds derive
deterministic IDs from source events, while query feeds assign random IDs when
records are created. Do not use event_id to infer event order.
Correlate a DNS change
Group operations records by customer, zone, and production_sha. Follow
change-approved, change-validated, and zone-change-applied to distinguish
authorization, validation, and successful deployment. Use the applied
serial to associate platform signing and propagation records.
Sort by source time while preserving those causal relationships. Knot may emit signing events during a transaction before the operations tool emits its successful completion event. Arrival order and timestamp order alone do not define the deployment sequence.
Map query detail
dns.queries.detail has a response schema with no log action. Use these
fields for response inspection:
| Fields | SIEM use |
|---|---|
event_id | Record identity and replay deduplication |
timestamp, server, customer_id, zone | Response time and serving context |
qname, qtype, rcode | Queried name, record type, and result |
query_transport, src_v4, src_v6 | Client address family and source |
ecs_v4, ecs_v6 | EDNS Client Subnet when supplied |
answer, answer_v4, answer_v6 | Returned answers |
The producer selects at most 1,000 detail events per second per sink using uniform reservoir sampling above that threshold. CNX then applies your subscription's sampling level before customer delivery: by default, up to 100 records per second across the fleet.
Request full fleet delivery from CNX if your consumer and SIEM can handle up to 10,000 records per second. This provides your full producer-selected stream, with producer sampling still in effect. See the query-detail reference for both sampling stages. Use detail for individual response inspection and sampled traffic analysis.
Customer delivery sampling applies only to query detail. Operations logs, platform logs, and query summaries have no customer sampling filter.
Map query summaries
dns.queries.summary contains aggregates rather than individual responses:
| Fields | SIEM use |
|---|---|
event_id | Record identity and replay deduplication |
timestamp, window_seconds | Window start and duration |
server, customer_id, zone, rcode | Grouping dimensions |
count | Classified response count for that group and window |
Sum count across servers for the same zone, response code, and window.
Summary counts are independent of detail sampling. All query delivery remains
best effort; interpret gaps using the
DNS delivery behavior.