Rebooting Microreboot: Architectural Support for Safe, Parallel Recovery in Microservice Systems

Laurent Bindschaedler

cs.DCApr 11, 2026

Rebooting Microreboot: Architectural Support for Safe, Parallel Recovery in Microservice Systems

Authors: Laurent Bindschaedler

Organizations: Max Planck Institute for Software Systems (MPI-SWS), Saarbrücken, Germany

Abstract

Microreboot enables fast recovery by restarting only the failing component, but in modern microservices naive restarts are unsafe: dense dependencies mean rebooting one service can disrupt many callers. Autonomous remediation agents compound this by actuating raw infrastructure commands without safety guarantees. We make microreboot practical by separating planning from actuation: a three-agent architecture (diagnosis, planning, verification) proposes typed remediation plans over a seven-action ISA with explicit side-effect semantics, and a small microkernel validates and executes each plan transactionally. Agents are explicitly untrusted; safety derives from the ISA and microkernel. To determine where restart is safe, we infer recovery boundaries online from distributed traces, computing minimal restart groups and ordering constraints. On industrial traces (Alibaba, Meta) and DeathStarBench with fault injection, recovery-group inference runs in 21 ms at P99; typed actuation reduces agent-caused harm by 95% in simulation and achieves 0% harm online. The primary value is safety, not speed: LLM inference overhead increases TTR for services with fast auto-restart.

Figures & tables

Panel of Figure 1 — Figure 1: System architecture. Telemetry feeds recovery-group inference. Agents propose remediation transactions in the Remediation ISA. The actuation microkernel (shaded) is the only trusted component that mutates infrastructure.

Action	Effect Type	Inverse/Comp.	Example Scenario
Restart	Restartable	Re-execute	Restart pod/service for transient failures
Drain	Compensatable	RestoreTraffic	Shield hub before disruptive changes
RestoreTraffic	Reversible	Drain	Resume traffic after planned drain
CircuitBreak	Reversible	Reset to default	Isolate unstable dependency
RateLimit	Reversible	Remove limit	Protect upstream from retry storms
Scale	Reversible	Scale by $-N$	Adjust capacity under resource stress

Table 1: Remediation ISA actions with effect types and semantics.

RQ	Experiment	Incidents
RQ2	Social Network harm validation (3 configs)	110
RQ2	Hotel Reservation harm validation	10
RQ3	TTR comparison vs Kubernetes auto-restart	15
RQ3	Multi-service parallel (5 services/incident)	3
RQ3	Reproducibility (3 seeds $\times$ 30)	90
RQ4	Fault diversity (5 types $\times$ 20)	100

Table 2: Online experiment incident counts.

	Simulation	Social Network	Hotel Res.
Configuration	(N=100)	(online)	(online)
Raw-Tools	77%	90% [74, 98] N=30	—
ISA-Only	24% ( $\downarrow$ 69%)	0% [0, 12] N=30	—
ISA+Critic	4% ( $\downarrow$ 95%)	0% [0, 7] N=50	0% [0, 31] N=10

Table 3: Harm rates by configuration and workload. Online columns show harm % with 95% exact binomial CI [ 8 ] .

Approach	Flexibility	Safety Mechanism	Rollback
Raw tools ( kubectl , APIs)	High	None	Manual
Policy verification	High	Post-hoc validation	Tool-specific
Undo mechanisms	High	State snapshots	Automatic
Typed ISA (ours)	Constrained	Effect-type enforcement	Transactional

Table 4: Comparison of safe actuation approaches.

Microservice Architecture

Microservice architecture organizes applications as independently deployable services and studies their design, operation, reliability, and coordination.

See all 19 papers

E2E-REME: Towards End-to-End Microservices Auto-Remediation via Experience-Simulation Reinforcement Fine-Tuning
Apr 13, 2026Lingzhe Zhang, Yunpeng Zhai, Tong Jia +5LLM-Based Program Synthesis Microservice Architecture
Peking University; Key Laboratory of Data Intelligence and Security Beijing, China · Alibaba Group China · Peking University; Key Laboratory of Data Space Technology and System Beijing, China
Safe Remediation as Risk-Constrained Intervention Decision in Microservice Systems
Jul 22, 2026Chengxiao Dai, Zhaokun Yan, Chenjun Lei +2Constrained RL
School of Computer Science, University of Sydney, Sydney, Australia · School of Computation, Information and Technology, Technical University of Munich, Munich, Germany · College of Business and Economics, Australian National University, Canberra, Australia +2
An Uncertainty-Aware Resilience Micro-Agent for Causal Observability in the Computing Continuum
May 11, 2026Suvi De Silva, Alfreds Lapkovskis, Alaa Saleh +2Agent Reliability Edge Computing
Department of Computer Systems and Sciences, Stockholm University, Stockholm 164 25, Sweden · Department of Computer Science, University of Helsinki, Helsinki 00014, Finland

Apr 13, 2026J/K move · Enter open · S save

E2E-REME: Towards End-to-End Microservices Auto-Remediation via Experience-Simulation Reinforcement Fine-Tuning

Lingzhe Zhang, Yunpeng Zhai, Tong Jia, Minghua He+4

Peking University; Key Laboratory of Data Intelligence and Security Beijing, China · Alibaba Group China · Peking University; Key Laboratory of Data Space Technology and System Beijing, China

LLM-Based Program Synthesis Microservice Architecture

Open paper AskPDF arXiv

Contemporary microservice systems continue to grow in scale and complexity, leading to increasingly frequent and costly failures. While recent LLM-based auto-remediation approaches have emerged, they primarily translate textual instructions into executable Ansible playbooks and rely on expert-crafted prompts, lacking runtime knowledge guidance and depending on large-scale general-purpose LLMs, which limits their accuracy and efficiency. We introduce End-to-End Microservice Remediation (E2E-MR), a new task that requires directly generating executable playbooks from diagnosis reports to autonomously restore faulty systems. To enable rigorous evaluation, we build MicroRemed, a benchmark that automates microservice deployment, failure injection, playbook execution, and post-repair verification. We further propose E2E-REME, an end-to-end auto-remediation model trained via experience-simulation reinforcement fine-tuning. Experiments on public and industrial microservice platforms, compared with nine representative LLMs, show that E2E-REME achieves superior accuracy and efficiency.

Abstract

Microservice Architecture

Explore similar work

E2E-REME: Towards End-to-End Microservices Auto-Remediation via Experience-Simulation Reinforcement Fine-Tuning

Safe Remediation as Risk-Constrained Intervention Decision in Microservice Systems

An Uncertainty-Aware Resilience Micro-Agent for Causal Observability in the Computing Continuum