Oct 16: DNS root key rollover breaks resolvers missing new anchor
Today's 3 things that matter
- DNS root key rollover breaks resolvers missing new anchor
RFC 5011 automated updates often fail on offline systems, read-only key stores, uninventoried embedded appliances, or outdated golden images; operators must verify key tag 38696 is in resolver config files and active in memory before the rollover takes effect. - Palo Alto patches PAN-OS root flaw under active exploitation
Network ops must prioritize patching PA-Series, VM-Series firewalls and Panorama immediately; Cloud NGFW and Prisma Access unaffected; limited but real exploit activity observed. - RPKI now covers 70% of routes after China's surge
Of 855,217 RPKI-covered IPv4 routes, 138,698 (16.2%) have a ROA looser than the route announced; operators should leave maxLength unset and publish ROAs for all announced prefixes to close attack windows.
Full stories below, grouped by topic.
Network Automation
Packet Pushers updates its open source network automation list
Packet Pushers · Oct 13, 2026 · Primary source
What happened: Packet Pushers maintains curated reference of open-source network automation tools. October 13, 2026 update documents maintenance status, recent activity, and adoption of Netmiko, Ansible, Nornir, Batfish, eNMS, and others.
Why it matters: Reference for practitioners evaluating which open-source automation tools are actively maintained and production-ready vs. stalled or deprecated; critical for avoiding build-on-abandoned-tools pitfall.
Packet Pushers maintains a living reference of open-source networking projects, last updated October 13, 2026. This curated list serves as a health check on the automation ecosystem: which projects have recent commits, active maintainers, and real operator adoption. Coverage includes configuration management (Ansible, Nornir, AVD), validation (Batfish, pyATS/Genie), device screen-scraping (Netmiko), and NMS platforms (eNMS). Tools like Netmiko show sustained activity and production adoption, while some community projects have been abandoned or absorbed into vendor offerings. For practitioners building new automation stacks, this reference answers critical questions: Is this tool maintained? Will it be supported in 2–3 years? Does the operator community still use it? The list reflects real-world adoption and vendor investment patterns, avoiding the trap of standardizing on deprecated tools.
Read the original at packetpushers.net
Data Center & AI Networking
Micas launches three 102.4T open Ethernet switches on Tomahawk 6
StorageNewsletter · Oct 15, 2026 · Industry news
What happened: Micas Networks announced three 102.4 Tbps open Ethernet switch platforms built on Broadcom Tomahawk 6, offering 800G air-cooled, 1.6T liquid-cooled, and 1.6T Davisson CPO variants. Portfolio enables practical migration from 800G to 1.6T while maintaining vendor-neutral architecture for mega-scale AI clusters.
Why it matters: ODM customers can deploy 1.6T Ethernet fabrics at scale without proprietary vendor dependencies; Tomahawk 6-Davisson CPO offers power-efficient path for cluster builders planning 2026-2027 mega-scale builds with 100K+ accelerators.
Micas' three-platform portfolio addresses different deployment profiles: air-cooled 51.2T Tomahawk 6 at 800G for hyperscalers with existing thermal budgets; liquid-cooled 102.4T variant with 1.6T pluggable connectivity for highest density; and 1RU 102.4T co-packaged system with 64 x 1.6T MMC DR8 optics for power-constrained mega-clusters. Each maintains vendor neutrality through Broadcom Tomahawk 6 architecture. Broadcom's VP of optical systems confirmed co-packaged optics will play increasing role as networks scale to higher bandwidth, highlighting Meta's deployment success with Tomahawk 5 'Bailly' CPO achieving 0.34% annual link failure rate—five times more reliable than traditional pluggable transceivers during first million hours of testing. Partners including Celestica, Delta, and Micas demonstrated these systems at OCP Summit (Oct 12-15), signaling rapid volume ramp. Support for Tomahawk 6's dual 100G/200G SerDes modes enables flexible capacity optimization.
Read the original at storagenewsletter.com
1.6T Ethernet samples now with volume ramp due in 2027
The Network DNA · Oct 16, 2026 · Analysis
What happened: Comprehensive technical roadmap of 1.6T Ethernet adoption covering IEEE 802.3dj standardization status, LPO/CPO architectures, optical module form factors (OSFP-XD, DR8, 2×FR4), deployment timeline, and practical guidance distinguishing 800G (mature, dominant) from 1.6T (early sampling, 2027 volume ramp).
Why it matters: Network engineers sizing 2026 builds should baseline 800G (cost-optimized, ubiquitous); validate 1.6T link budgets for spine/inter-cluster links targeting 2027 adoption; prioritize OSFP-XD form-factor qualification to avoid proprietary module lock-in.
800G remains the dominant deployed speed across hyperscale AI fabrics and is the natural upgrade path for most large-scale builds in 2026, while 1.6T is entering early sampling with hyperscalers and neoclouds but lacks ubiquity. IEEE 802.3dj targets completion in late 2026, enabling 1.6T-capable NICs to ship with next-generation accelerators in Q4 2026 and Q1 2027. Optical landscape includes 1.6T DR8 (500m O-band, 22W typical), DR8-2 (2km), 2×FR4 de-facto standard using CWDM (2km reach), SR8 multimode (50-100m 850nm), and coherent variants (1600ZR/ZR+ to 1,000km, 32-40W power). OSFP-XD with dual MPO-12 APC or single MPO-16 APC enables 16-fiber connectivity per 1.6T link. Module pricing is expected to decline from current $2,500-3,500 to $1,500-2,000 by late 2026, achieving cost-per-bit parity with 800G by 2027. Practitioners should validate backplane electrical-interface (1.6TAUI-8/-16) compatibility and firmware support before committing to 1.6T; multi-vendor interoperability remains nascent and platform-specific.
Read the original at thenetworkdna.com
Routing & Internet
RPKI now covers 70% of routes after China's surge
IPinfo / daily.dev · Oct 16, 2026 · Industry news
What happened: RPKI now secures 70.3% of global Internet routes, up from 54.0% a year earlier, with China's adoption leap (3.7% to 83.0%) driving most of the gain. However, RPKI only validates origin, not path: a August 2026 Hetzner hijack exploited a loosely scoped ROA to forge a route and serve malware, and roughly 16-21% of covered routes still have such 'open windows.'
Why it matters: Of 855,217 RPKI-covered IPv4 routes, 138,698 (16.2%) have a ROA looser than the route announced; operators should leave maxLength unset and publish ROAs for all announced prefixes to close attack windows.
RPKI-valid routes cover 70.3% of routed prefixes, up from 54.0% a year earlier, with IPv4 addresses rising from 53.1% to 65.0%. Much of the jump came from China moving from 3.7% to 83.0% of its IPv4 addresses in valid routes, accounting for nearly 60% of the global gain. However, RPKI validates only the origin AS, not the path taken by the route, and cannot catch attackers who stay within published ROA constraints. Roughly 16.2% of IPv4 and 20.9% of IPv6 RPKI-covered routes have loose 'open windows'—ROAs broader than the routes announced, creating gaps exploitable by forged-origin hijacks. The August 2026 Hetzner incident demonstrated this: attackers exploited a loosely scoped ROA authorizing origination at /24 lengths within a /16 block, received RPKI-valid validation, and used the route to obtain a TLS certificate and serve malicious updates. Operators should publish ROAs for everything announced (three in ten routed prefixes remain not RPKI-valid), avoid setting maxLength (one in six covered IPv4 routes has a gap like Hetzner's), and verify ROAs are still present, as several operators including Saudi Telecom, SEACOM, and Orange Maroc have recently lost some of theirs.
Read the original at ipinfo.io
DNS root key rollover breaks resolvers missing new anchor
Akamai · Oct 11, 2026 · Industry news
What happened: On October 11, 2026, the DNS root zone switched to signing with KSK-2024 (key tag 38696), replacing KSK-2017. Any validating resolver lacking this trust anchor fails to resolve both signed and unsigned domains, causing widespread SERVFAIL errors within 48 hours.
Why it matters: RFC 5011 automated updates often fail on offline systems, read-only key stores, uninventoried embedded appliances, or outdated golden images; operators must verify key tag 38696 is in resolver config files and active in memory before the rollover takes effect.
On October 11, 2026, the DNS root zone stopped signing with KSK-2017 (key tag 20326, active since 2018) and switched to KSK-2024 (key tag 38696), only the second KSK change in DNSSEC history. Resolvers can cache the old key set for up to 48 hours, so failures manifest within two days as the cache expires. KSK-2024 was published in the root zone on January 11, 2025, giving resolvers 21 months to pick it up either automatically through RFC 5011 (watching the root for new keys and trusting one after 30 days residence) or through software updates. Both keys use RSA/SHA-256, so no algorithm change occurred. Operators running DNSSEC-validating resolvers must ensure their systems trust key 38696 beforehand; a missing trust anchor breaks DNS broadly across an entire resolver, while zone-specific DNSSEC issues affect individual domains. Organizations should verify their baseline DNS health before October 11 to rapidly distinguish local configuration issues from root key rollover outages during troubleshooting.
Read the original at akamai.com
Network Security
Palo Alto patches PAN-OS root flaw under active exploitation
ITdaily · Oct 15, 2026 · Industry news
What happened: Palo Alto Networks disclosed CVE-2026-0322, a critical vulnerability in PAN-OS management interface allowing authenticated attackers with network access to execute arbitrary code with root privileges. CISA added it to Known Exploited Vulnerabilities catalog with evidence of active exploitation.
Why it matters: Network ops must prioritize patching PA-Series, VM-Series firewalls and Panorama immediately; Cloud NGFW and Prisma Access unaffected; limited but real exploit activity observed.
Palo Alto Networks disclosed CVE-2026-0322 in PAN-OS on October 14, with a CVSS score of 8.6 and high severity rating. The flaw allows an authenticated attacker with network access to execute arbitrary code with root privileges on the firewall, affecting PA-Series and VM-Series firewalls and Panorama. Very limited exploitation targeting the management interface has been observed, with three IP addresses tied to exploit attempts. CISA added CVE-2026-0322 to its Known Exploited Vulnerabilities catalog and gave federal civilian agencies three days until October 17, 2026 to remediate. Unlike the GlobalProtect VPN bypass vulnerabilities, this requires authentication but affects the most critical management plane. Organizations with internet-exposed management interfaces or those allowing authenticated remote access face immediate risk.
Read the original at itdaily.com
FBI says FortiBleed has harvested 86,644 Fortinet credentials
The Hacker News · Oct 9, 2026 · Industry news
What happened: FBI and Secret Service issued a joint advisory warning that the FortiBleed credential harvesting campaign remains active, with over 86,644 compromised FortiGate firewall and SSL VPN credentials verified across 194 countries. The campaign exploits reused credentials and legacy SHA-256 password storage for GPU-accelerated cracking.
Why it matters: Active threat to internet-facing FortiGate appliances with verified credential databases in use by organized actors; requires immediate detection of unauthorized access and credential rotation across all FortiOS versions.
The FBI and Secret Service warned that the FortiBleed credential harvesting campaign remains an active threat aimed at internet-facing Fortinet FortiGate firewalls and SSL VPN gateways. The campaign exploits reused or leaked credentials and legacy SHA-256 password storage, enabling threat actors to harvest and crack authentication data at scale, with attackers continuing to scan internet-exposed Fortinet firewalls using previously obtained compromised credentials. The Russian-speaking operation is estimated to have netted more than 86,644 working device credentials spanning 194 countries as of June 19, 2026. In some cases, criminals are creating their own administrator accounts and changing or deleting legitimate accounts, leaving companies unable to access their own devices. Unlike zero-day vulnerabilities, FortiBleed relies on credential stuffing against exposed management interfaces and weak password practices. Network defenders must audit FortiGate access logs for anomalous logins, implement IP-based access restrictions, and force credential rotation immediately.
Read the original at thehackernews.com
Agentic AI & MCP
NetBrain R13 adds agent team and MCP server
NetBrain Exchange Community · Oct 13, 2026 · Primary source
What happened: NetBrain R13 release webinar on October 13, 2026 covering the Agent Team (Diagnosis, Assessment, Change, Companion agents) working with the NetOps Harness and MCP Server. Demonstrates end-to-end network change validation and CVE exposure detection.
Why it matters: Shows production agentic NetOps with MCP as the interop layer; Assessment Agent identifies security exposure proactively, Change Agent validates modifications before execution.
The October 13 webinar announced the NetBrain R13 release, available in early October 2026 to all customers. The core innovation is the Agent Team + NetOps Harness architecture: a unified platform where Diagnosis, Assessment, Change, and Companion agents operate on live network context with built-in guardrails. The Assessment Agent uses the Golden Assessment Library—pre-built network automations derived from NetBrain's customer base—to surface CVE exposure and configuration drift before they become incidents. The Change Agent plans, validates, and documents changes end-to-end, reducing change execution and validation effort by 85% according to internal data. Critical for operations teams: MCP Server exposes NetBrain network intelligence (topology, path calculations, automation tools) to external agents from AI SRE platforms (Resolve AI, Azure SRE Agent), AIOps (BigPanda), ITSM (ServiceNow), and custom agents built with Claude or OpenAI models. This bi-directional MCP integration positions the NetOps Harness as the network context layer in broader IT Ops ecosystems—agents from incident management, SRE, and ITSM can query live network state and trigger network-aware workflows without reimplementing network topology logic.
Read the original at exchange.netbrain.com
Research, Standards & Industry
OpenTelemetry Java agent 3.0 reaches release candidate
OpenTelemetry Blog · Oct 13, 2026 · Primary source
What happened: OpenTelemetry Java agent v2.32.0 released as RC for 3.0 targeting October 2026; Collector Contrib v0.160.0 adds RC process semantic conventions metrics to hostmetrics receiver, advancing stable v1 release candidate maturity.
Why it matters: Java shops can preview OTel 3.0 behavior before GA; ops teams gain standardized process-level metrics (CPU, memory, state, timing) for infrastructure observability, unlocking AIOps anomaly detection on system resource consumption.
The OpenTelemetry Java agent 2.32.0 serves as the release candidate for 3.0, allowing early adopters to validate changes before the final GA release targeted for October 2026. Alongside this, OpenTelemetry Collector Contrib v0.160.0 introduces release candidate support for process semantic conventions metrics in the hostmetrics receiver's process scraper, marking the final stage before stable v1 status. These metrics include process status, CPU usage, memory footprint, and lifecycle timestamps—critical signals for AIOps systems to correlate application behavior with host-level resource contention. For SRE and network operations teams, this enables unified observability pipelines where process metrics (e.g., network daemon CPU burn, DNS resolver memory growth) flow through the same Collector infrastructure as network telemetry, allowing correlation-based root cause analysis. The semantic conventions standardization ensures metrics naming and cardinality consistency across vendors and languages, reducing the friction of multi-language observability integration common in cloud-native networks.
Read the original at opentelemetry.io
IETF draft ties segment routing SIDs to guaranteed resources
IETF Datatracker · Oct 13, 2026 · Primary source
What happened: Active IETF draft (v18, May 2026 publication, expires November 2026) describing mechanisms to allocate network resources (bandwidth, compute, memory) to Segment Routing Identifiers (SIDs) for both SR-MPLS and SRv6, enabling resource-guaranteed policies and Network Resource Partitions.
Why it matters: Network operators can now engineer SR policies with explicit resource reservations without RSVP-TE overhead, enabling deterministic slicing for 5G/6G enhanced VPN services and compute-aware routing without per-path state explosion.
The draft-ietf-spring-resource-aware-segments working group document (v18, expires Nov 16, 2026) defines how to associate network resource subsets (bandwidth pools, dedicated processor time, memory buffers) with individual SIDs or groups of SIDs. This extends SR beyond pure topology steering to include resource-aware path selection: a resource-aware SID retains its forwarding semantics but also identifies the set of underlay resources reserved for packets using that SID. The mechanism supports centralized controller-driven resource planning based on service traffic matrices and performance requirements, addressing a long-standing gap where SR policies lacked the resource guarantee capabilities of RSVP-TE. Implementations are already in production on Huawei ATN9XX and CX600 routers. For network architects, this enables declarative service slicing: define an enhanced VPN intent ("latency < 10ms, throughput > 10Gbps") and map it to resource-aware SR policies, with the controller allocating appropriate SID pools and advertising them via BGP-LS. The draft remains active; standards track completion is expected soon, making this practical for near-term deployment in tier-1 carrier and cloud networks planning deterministic slices.
Read the original at datatracker.ietf.org
AI Model Providers
Google ships Gemini 4 Argon with 1M token context
Google Developers Blog · Oct 13, 2026 · Primary source
What happened: Gemini 4 Argon became globally available October 13, 2026, through the Gemini API with a 1.0M token context window and up to 1M output tokens, designed for complex coding, agentic tasks, mathematics, science, and cybersecurity workflows. Introductory pricing is $2 per million input tokens and $10 per million output tokens through November 30, then $4/$20.
Why it matters: Gemini 4 Argon is optimized for long-horizon workflows requiring deep reasoning and structural precision, including code migrations and infrastructure automation, making it relevant for AIOps and complex system design.
Gemini 4 Argon marks a step change in physical understanding, code generation, and visual-semantic reasoning, unlocking new workflows across the robotics stack and developer tooling. The rollout coincided with broader Gemini 4 Argon integration into Android Studio's developer environments on October 13, 2026. Developers can access Argon through Google AI Studio, the Gemini API, and Google Antigravity to build applications leveraging simulation tools like MuJoCo and code-as-policies for seamless transitions from real-world data to physically verified control and self-improvement workflows. Google made it globally available on October 13, 2026 at 16:00 UTC with Pro and Ultra tier subscribers gaining immediate access in the Gemini app. The aggressive introductory pricing and broad API availability position this as a direct competitor for infrastructure automation coding tasks, particularly relevant for teams evaluating reasoning models for complex orchestration and policy generation in robotics and physical systems.
Read the original at blog.google
xAI's Grok Voice 3.0 tops speech-to-speech benchmark
xAI · Oct 14, 2026 · Primary source
What happened: xAI announced Grok Voice 3.0 on October 14, 2026, a speech-to-speech model available in the Voice Agent API at $0.08 per minute with sub-one-second latency. Artificial Analysis ranks it first on the Speech-to-Speech Index at 86.9, ahead of GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6.
Why it matters: Grok Voice 3.0 powers millions of conversations daily in Tesla and SpaceX infrastructure, handling 20,000 Starlink support calls daily with 70% fully automated, critical for evaluating production voice-driven AIOps automation and agent infrastructure.
xAI released Grok Voice 3.0 on October 14, 2026, eleven weeks after Grok Voice 2.0; the grok-voice-latest alias immediately moved to 3.0 with Grok Voice 2.0 marked deprecated. Artificial Analysis testing dated October 13 reported 98.5% on Big Bench Audio speech reasoning, 94.8% task success rate, and 97.0% on Full Duplex Bench for natural turn-taking and interruption handling. Pricing remained unchanged at $0.08 per minute. Critical operational detail: users running the grok-voice-latest alias received automatic model upgrades without code deployment on October 14, creating a silent production behavior change for existing integrations. The WebSocket endpoint, session parameters, and events remained stable. The production workload scale (Tesla, SpaceX, Starlink) and unchanged pricing make this highly relevant for operators evaluating voice-driven automation at scale in production SRE workflows.
AI Industry & Policy
US and China models leave middle powers behind
Fortune · Oct 14, 2026 · Analysis
What happened: US and China dominate AI model development with 59 and 35 notable models respectively through 2025; only South Korea has multiple notable models among other nations. Private sector actors including Nvidia and Anthropic have become geopolitical actors, marginalizing middle powers from the AI field.
Why it matters: For enterprise teams planning AI infrastructure investments, concentration of development and resources in US/China firms shapes vendor ecosystem and supply-chain dependencies for mission-critical workloads.
Stanford University's 2026 AI Index reveals stark concentration in model development: the U.S. led with 59 notable AI models (by citation rate, compute costs, and usage levels) while China produced 35; South Korea was the only other country with multiple notable models, at eight. The analysis highlights that private sector actors—particularly Nvidia and large AI labs—have stepped into roles traditionally held by national intelligence and defense agencies, making private companies de facto geopolitical actors. For middle powers, this concentration represents a structural disadvantage: they play only tangentially in the AI field despite its expanding global influence. The implication for enterprise infrastructure teams is significant: AI capability, model availability, and supply-chain dependencies are increasingly controlled by a handful of jurisdictions and private actors, affecting choices around cloud providers, chip sourcing, and model selection.
Read the original at fortune.com
Canada adds $500M to regional AI initiative
Canadian Manufacturing · Oct 13, 2026 · Primary source
What happened: Canada's Minister of AI announced $500M in additional federal funding over five years for the Regional Artificial Intelligence Initiative, bringing total federal AI funding to $700M. The investment supports the National Artificial Intelligence Strategy announced in June 2026 to strengthen AI capabilities and ensure responsible adoption.
Why it matters: Signals how governments are operationalizing AI adoption strategies with concrete funding; relevant for enterprises planning Canadian operations and understanding government incentives for regional AI deployment.
On October 13, Minister Evan Solomon announced $500M in new funding, expanding total federal support for the Regional AI Initiative to $700M. This initiative is part of Canada's broader National Artificial Intelligence Strategy (announced June 2026), designed to strengthen domestic AI capabilities and ensure responsible, country-wide AI adoption. The timing and scale suggest a shift from exploratory AI pilots to structured, government-backed scaling efforts. For enterprises with Canadian operations or partnerships, this investment structure indicates growing government coordination with private sector adoption targets and regulatory frameworks. The initiative likely includes regulatory sandboxes and incentive structures that affect how organizations deploy AI tools in regulated sectors (energy, healthcare, finance) where provincial oversight intersects with federal strategy.
Read the original at canadianmanufacturing.com
Read this edition on the web · The week in network intelligence · Vendor Radar
Digital Plumber is AI-curated and AI-summarized, with no human review before publishing. Verify before acting on anything here. How it works.