Astragalus: Automatic Configuration Repair for Production Networks DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 403

Zhenrong Gu, Peng Zhang, Xing Feng, and Xu Liu · 2026

Read Astragalus: Automatic Configuration Repair for Production Networks DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 403 by Zhenrong Gu, Peng Zhang, Xing Feng, and Xu Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-22092.txt.

Introduction

Zhenrong Gu and co-authors present Astragalus, a tool for automatic configuration repair in production networks. Configuration errors can cause catastrophic service outages, and existing repair tools follow a semantics-driven approach, modeling network semantics as SMT constraints, solved for an error location or fix. Building and solving such constraints is complex, so Astragalus differs, repairing every incident across several sizes of a synthesized network and 97.5 percent of incidents on a real network, with 15 error types, in an average of 6.93 seconds, plus valid repairs under 6 minutes for 7 recent real incidents.

Catalog record

CREATORZhenrong Gu, Peng Zhang, Xing Feng, and Xu Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59820
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.22092

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Astragalus: Automatic Configuration Repair for Production Networks DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 403 Zhenrong Gu, Peng Zhang, Xing Feng, and Xu Liu Abstract . Network configurations are prone to errors, which can lead to catastrophic service outages. A tool that can achieve automatic configuration repair (ACR) is highly desired by operators. Existing tools for ACR follow a semantic-driven approach : they model network semantics as a set of SMT constraints, and solve them for a location or fix of the error. Due to the complex semantics of networks, constructing and solving these constraints can be prohibitively expensive, making these tools neither general nor scalable. Inspired by automatic program repair (APR), we explore another direction, i.e., a syntax-driven approach , which tries to repair program bugs by “grafting” some existing code in the same repository, without modeling program semantics. Following this direction, we propose Astragalus, a syntax-driven method for ACR. It uses multiple iterations of a “localize-fix-validate” pipeline to search for repairs, and proves quite effective on configurations of our production network. Specifically, we show that Astragalus can repair every incident in multiple sizes of a synthesized network, and 97.5% of the incidents on a real network, both with 15 types of errors injected, within an average time of 7.36 seconds. It has also provided valid repair suggestions in under 6 minutes for 4 recent network incidents or undesired changes, in a real production network with O(1,000) ∼ \sim O(10,000) devices. 1. Introduction Network outages frequently make headlines due to misconfigurations, software bugs, and hardware failures ( 28 ) . Among them, misconfiguration is a major source of network outages ( 20 ; 32 ) . In the network of a large cloud service provider (CSP, referred to as Provider Alater), misconfiguration account for 35.4% of all incidents in the past 5 years. To detect network misconfigurations, both the academia and industry have proposed a lot of network verifiers ( 12 ; 34 ; 33 ; 1 ; 16 ; 11 ; 8 ; 31 ) . These tools can check whether network forwarding behaviors align with operator intent, and conform to properties like loop freedom, and blackhole freedom. However, network verifiers can only detect issues or reduce the impact rather than resolve it, and operators still need to manually localize the root causes and repair the configurations. Such a manual process can be frustrating and time-consuming, thereby delaying planned network updates. Taking Provider A for example, the verification mostly finish within 3 minutes, while operators take more than 10 minutes to localize and repair 45.8\% of the cases, with the longest case taking 5 hours. Similar to the Automatic Program Repair (APR) tools that greatly facilitate software development ( 13 ; 22 ) , a tool that can achieve Automatic Configuration Repair (ACR) is highly desired by operators to help them quickly localize the root causes and resolve the incidents. Towards ACR, there have been a handful of excellent tools ( 14 ; 2 ; 15 ) , which can localize configuration errors or generate repair for misconfigurations. These tools encode network semantics, forwarding properties, and configurations as a set of SMT constraints, and solve them with off-the-shelf solvers, e.g., Z3. However, we find them still limited generality and scalability. This is mainly due to the network semantics, especially the distributed routing protocols, is very complex. As a result, the constraints, even already very complex, can only cover a subtset of protocol or features, so may not be general enough to be applied to production networks. In addition, the complexity of routing protocols also makes the number of constraints explode with the network size. For example, we find that for a fattree network of 45 nodes running BGP, the number of SMT constraints is already as large as 7900. Such a large number of constraints make it prohibitively expensive for SMT solvers to solve in a reasonable time. Before diving into the design of ACR, let us first visit APR for some inspiration. In APR, there are broadly two classes of approaches, i.e., the semantics-driven approaches and the syntax-driven approaches. The semantics-driven approaches, or correct-by-construction approaches, formally encode the program repair problem as a set of SMT constraints, and solve them for a repair ( 23 ; 21 ) . The syntax-driven approaches, or generate-and-validate approaches, leverage the “plastic surgery hypothesis” ( 6 ) : the code ingredients that can repair the software already exist somewhere in the codebase. These approaches generate a set of candidate updates, and validate them to find a repair, Without modeling program semantics ( 18 ; 19 ; 26 ) . We can seen that existing ACR methods e.g., ( 14 ; 2 ; 15 ) , fall within the first category of “semantics-driven” approaches , and acknowledge the scalability and generality limitations. In this paper, we explore the other direction, i.e., the “syntax-based” approach, and find it a quite promising approach for ACR, due to the following reasons. First, the “plastic surgery hypothesis” also holds for network configurations, since multiple devices often take the same role with similar configurations, and misconfigured devices can be repaired based on configuration lines from other devices of the same role. Second, for syntax-driven approaches in APR, the validation process often dominates the total repair time, due to recompiling and rerunning parts or the entire program ( 27 ) . For network, configuration verifiers are already very fast and it is possible to validate a large number of candidate fixes for high repair ratio. Following the direction of syntax-driven approach, we propose Astragalus, a general and scalable tool that can achieve ACR for large-scale production networks. Astragalus consists of three steps: (1) Localization. We localize suspicious configuration lines with techniques from APR, Spectrum-Based Fault Localization (SBFL) ( 4 ; 17 ) . The intuition for these techniques is that faulty lines contribute more to property violations than correct ones. (2) Fix Generation. According to the suspiciousness of configuration lines, we generate a set of candidate fixes including remove, insert, and modify, based on existing configuration in the same network. (3) Validation. We use off-the-shelf configuration verifiers to check whether the update resolves the violation without any side effects. In summary, our contribution includes: • We motivate a shift from semantics-driven paradigm to a syntax-driven paradigm for ACR, based on some insights from program repair. • We design and implement Astragalus to realize ACR, a syntax-driven approach that is more scalable and general than existing methods based on SMT models. • We test Astragalus on both real faults and synthesized network configurations with 15 types of injected errors, and show that it can repair all synthesized errors of O(100) real devices with an average time of 6.97 seconds. 2. Motivation In this section, we discuss the hurdle of current configuration repair approaches to motivate the need for a syntax-driven approach to ACR. Then, we contrast network configurations with software code to introduce the challenges of a syntax-driven approach. 2.1. Semantics-driven repair Existing work has explored synthesis-based methods for root-cause diagnosis and automated repair. Synthesis methods build a formal model of the network

Read the complete text