• /
  • EnglishEspañolFrançais日本語한국어Português
  • ログイン今すぐ開始

Autopilot 2.7

November 10, 2026

What's new

Troubleshoot Kafka with Autopilot

Autopilot can now better troubleshoot Kafka issues and recommend next steps to mitigate them, for Kafka monitored with the New Relic on-host integration, OpenTelemetry, Amazon MSK, and Confluent Cloud.

  • Cluster health: Checks consumer lag, under-replicated and offline partitions, and broker and controller state.
  • Entity relationship mapping: Maps relationships across the ecosystem, connecting producers, consumers, topics, brokers, clusters, and host infrastructure.
  • Client performance: Reviews a producer's or consumer's request and fetches latency, I/O wait, consume rate, and poll timing.

Change Tracking for GitHub Copilot

preview

We're still working on this feature, but we'd love for you to try it out! This feature is currently provided as part of a preview program pursuant to our pre-release policies.

Autopilot can now help monitor your service after each deployment and report the results on your pull request. At six checkpoints after a deployment (10 minutes, 30 minutes, 1 hour, 2 hours, 24 hours, and 72 hours), Autopilot compares post-deployment metrics against a baseline and posts the result as a PR comment. When it detects a regression, Autopilot posts its findings to Slack and opens a GitHub issue assigned to GitHub Copilot. With GitHub Copilot Enterprise, Copilot can propose a fix as a pull request.

  • Checkpoint reports on your PR: At each checkpoint, Autopilot comments on the pull request with the result — no regression, inconclusive, or regression detected — so your team sees the signal without leaving GitHub.
  • Slack notifications: If you've connected Slack, Autopilot posts a message when monitoring starts for a deployment and replies in the same thread at each checkpoint.
  • Automatic regression issues: When Autopilot detects a regression, it opens a GitHub issue with what regressed, the supporting evidence, where to look, and acceptance criteria, and assigns the issue to GitHub Copilot.
  • PR label trigger: Add the watchpoint_monitoring label to a pull request before you deploy to turn on monitoring for that deployment, with no need to set the watchpoint_monitoring attribute in your Change Tracking event. You choose which deployments to monitor, with no pipeline changes required.

For setup instructions and full prerequisites, see Monitor deployments with New Relic Change Tracking for GitHub Copilot.

Kafka troubleshooting, deployment monitoring, and more

New features

  • Troubleshoot Kafka with Autopilot: Autopilot can now better troubleshoot Kafka issues and recommend next steps to mitigate them, for Kafka monitored with the New Relic on-host integration, OpenTelemetry, Amazon MSK, and Confluent Cloud.
  • Change Tracking for GitHub Copilot (Preview): Autopilot can now help monitor your service after each deployment and report the results on your pull request.
  • View Autopilot usage data by user: Using Consumption Management you can now see Autopilot consumption usage broken down by user.
  • Logs and metric discovery for infrastructure: During infrastructure investigations, Autopilot can now review an entity's logs and look up the metric names it reports.

Improvements & fixes

  • Improved issue search: Autopilot now resolves alert issues more reliably — suffix name variants no longer hide behind exact matches, closed issues are more reliably resolvable for retrospective investigation, and reporting status is surfaced so stale entities don't win over live ones.
  • More complete quick answers: Lists show each item, up to 50, and answers include details behind counts, like critical entities and open issues.
  • CCU totals: CCU totals in quick answers no longer include data volume figures.
  • Time ranges you ask for: Autopilot may ask you for a time range for investigation.
  • Focused checks on ongoing issues: Asking whether a problem is still happening now checks the most recent period.
  • Improved incident start times: Incident investigations now start from the time the issue opened.
  • More reliable question routing: Autopilot has improved path handling based on the context of your question.
  • Investigations continue when a tool service is down: Autopilot continues with the remaining tools when a connected tool service is unavailable.
  • Findings kept when an investigation reaches its budget: Autopilot can now summarize the findings it gathered before an investigation reached its budget limit.
  • More resilient investigations: Autopilot is more likely to complete an investigation when one of its connected tool services is unavailable, and better able to recover from temporary slowdowns.
  • Fuller investigation summaries: Summaries of multi-step investigations are more likely to include the full analysis, and charts appear in more of them.
  • Better answers for Synthetics questions: Autopilot can now better answer questions about the lists, counts, and status of all monitors in an account, and better resolves an account you name in your question. Analysis of a failed check now focuses on the step that failed.
  • Saved preferences fit the current question: Preferences saved from earlier conversations are less likely to over influence what your current question needs, such as a clarifying question when no entity is named.
  • More from your knowledge base collections: When you specify a knowledge base collection, Autopilot now includes more of its results and is more likely to call out if it cannot retrieve a document.
  • More alert context in investigations: Autopilot can now draw on more alert condition detail, such as thresholds and the affected entity, when investigating an issue.
  • Workload analysis: Autopilot can now review the member entities of a workload and count entities by attribute.
  • Consistent timestamps in quick answers: Timestamps and durations in quick answers are more consistent.
  • Incident lookups for open issues: Resolved an issue that caused Autopilot to report no incidents for a valid, open issue.
  • Accurate step status: Resolved an issue that caused an unfinished investigation step to show as complete.
  • Entity search reliability: Resolved an issue that caused some entity searches to not trigger.
  • New conversations from the alerts page: The New conversation button now opens a new thread from the alerts page.
  • Page-specific example prompts: Pages that provide their own example prompts now show them in the Autopilot panel.
Copyright © 2026 New Relic株式会社。

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.