DNS resolvers are designed for continuous operation and typically do not require regular restarts. In real-world environments, however, restarts are sometimes unavoidable. Configuration changes, software updates and security patches can make a restart necessary. What is often underestimated is the operational impact such a restart can have on performance and stability – especially in large-scale DNS environments.
When a resolver restarts, it loses its entire runtime state. This includes the recursive DNS cache with all previously resolved records, negative responses such as NXDOMAIN entries, DNSSEC validation state, runtime statistics, and intelligence about authoritative name server performance. In effect, the resolver forgets everything it has learned. All previously answered queries must be resolved recursively again. In smaller environments, this may result only in a brief warm-up phase. In larger deployments with many resolver instances and high query volumes, however, latency and CPU load can increase noticeably. At the same time, traffic to root and TLD servers rises as queries must be repeated.
Particularly critical is the loss of accumulated knowledge about authoritative servers. During normal operation, resolvers continuously build detailed performance profiles, tracking reachability, response times, timeout behavior, EDNS capabilities, cookie support, and overall reliability. This intelligence directly influences which authoritative servers are preferred for future queries. When that data is lost, the resolver must reassess and benchmark servers from scratch. Until this performance intelligence is rebuilt, higher latency, additional retries, and less stable resolution behavior may occur.
PowerDNS Recursor Plexus addresses precisely this challenge. As a proprietary add-on for PowerDNS Recursor, Plexus prevents the loss of critical operational data during restarts by enabling state sharing across multiple Recursor instances. Instead of operating as isolated nodes, resolvers become part of a coordinated mesh powered by NATS messaging – a technology that has already proven its effectiveness in our DNSdist Defender add-on. When a Recursor instance restarts, it does not begin with an empty cache and no context. Instead, it is repopulated with relevant state information from its recursive peers through Plexus, preserving operational continuity and significantly reducing post-restart performance degradation.
In this first release, the data collected and distributed by Plexus includes recursive cache contents such as record cache entries and DNSSEC validation state. In addition, Plexus shares authoritative server status information in the form of an authoritative NS speed table. This table contains response time metrics, timeout history, historical failure patterns, and overall server reliability data. As a result, a restarted resolver immediately regains the necessary insight to prioritize the fastest and most reliable authoritative servers, without having to retest them. This directly reduces latency, minimizes timeouts, and stabilizes overall resolution performance.
PowerDNS Recursor Plexus delivers particular value in large enterprise and service provider environments. In infrastructures with numerous PowerDNS Recursor instances, high query rates, and strict latency requirements, the cumulative effects of restarts can be substantial. Without Plexus, each restarted instance must rebuild its cache and performance intelligence from scratch. With Plexus, operational knowledge is preserved, maintenance windows can be executed without significant performance penalties, and DNS infrastructure remains stable and responsive – even during restarts.
If Recursor restarts and prolonged warm-up phases are impacting your performance, get in touch for a quote. We’ll show you how PowerDNS Recursor Plexus can keep your DNS infrastructure fast, stable, and resilient.
