LEARN. PRACTICE. GROW.Reconstruct the client path
A fictional operations workstation needs configuration, name resolution, and a connection to the service. Retain separate evidence for each stage: address and lease, gateway, resolver used, DNS question, answer, TTL, and connection attempt. If some clients work and others fail, compare their state instead of concluding that the whole network is healthy. One client can retain a valid lease while another needs a new allocation. One resolver can retain old data while another queries authority after a change.
Read the DHCP timeline
For an eight-hour lease starting at 09:00, the model uses T1=50% and T2=87.5% without random variation. Renewal begins at 13:00, rebinding at 16:00, and the lease expires at 17:00. These times are relative to the start rather than added sequentially. Before expiry, renewal failure does not imply immediate loss of a still-valid address. If the lifetime ends without DHCPACK, the client can no longer use the former lease. In production, inspect received values instead of assuming every server supplies defaults.
Follow relay and count the pool
A relay can serve clients on another subnet, but the server needs applicable context and configuration. If giaddr identifies 10.72.40.1, check scope and policy for the client network; receiving DISCOVER does not prove an eligible address exists. In the model with 100 addresses, 12 exclusions, eight separate reservations, and 65 dynamic leases outside those sets, 15 remain. Document overlaps before subtraction. Two offers with different gateways also require source and authorization checks; the fastest offer is not automatically correct.
Distinguish negative DNS responses
NXDOMAIN and absence of one record type are not the same observation. Without CNAME, an authoritative NOERROR response lacking AAAA and containing SOA can indicate NODATA for that type while the name has A. Record name, type, class, and flags before reporting that the domain does not exist. SERVFAIL represents another situation requiring its own hypotheses. At shift handover, supply the observed question and answer, resolver, and time. A description such as DNS broken discards information useful for choosing the next test.
Calculate validity without restarting the clock
In the negative-cache exercise, SOA TTL=600 and SOA.MINIMUM=300 produce an initial 300-second bound. After 120 seconds, 180 remain, excluding prefetch, serve-stale, and local adjustments. Querying the copy does not automatically restart its lifetime. If the name is created meanwhile, a consumer can still observe the former negative response during its remaining validity. The operational decision must account for that consumer, not merely demonstrate that authority now knows the name. Controlled clearing can be a local measure without promising effects on other resolvers.
Plan a change before cutover
Lowering TTL at 22:00 does not retroactively change a response received at 21:59 with a one-hour lifetime. That copy can remain valid until 22:59 under the model assumptions. Plan reduction ahead of time and inspect consumers, caches, and connection pools. Maintain approved compatibility or coexistence while old consumers remain. DNS lifetime also does not necessarily bound an already-open connection. Define functional cutover and rollback criteria for relevant file jobs and integrations rather than only for a manual query.
Test transport and close with evidence
A UDP response with TC=1 can lead to a new DNS attempt over TCP. If the firewall denies TCP/53 to the authorized resolver, small queries can work while others fail. Correlate capture and rule before changing the required path. Do not use a ping reply as proof of functioning DNS or DHCP. After correction, repeat the failing operation and check received parameters, correct resolution, and the complete connection. Record eliminated hypotheses and test limits, including clients or resolvers not observed.
Lease 09:00–17:00: T1 at 13:00 and T2 at 16:00 in the model. A 300-second negative cache has 180 remaining after 120.
Common pitfalls
Restarting lifetimes on each query, treating new TTL as retroactive, or accepting a running process as a valid allocation.
Related topics: DNS and DHCP · Change management
Stored state and time explain differences between clients using the same infrastructure.
Reference: Dynamic Host Configuration Protocol · N10-009 V9