alibabacloud-sysom-diagnosis（阿里巴巴云系统诊断）

SysOM Diagnosis (sysom-diagnosis)

Run diagnostics with ./scripts/osops.sh under sysom-diagnosis/ (skill root). stdout is a JSON envelope. By default it performs a local quick check without cloud invocation; remote deep diagnosis is triggered by --deep-diagnosis (memory domain) or io/net/load subcommands, with built-in environment checks.

Installation

Pre-check: Aliyun CLI >= 3.3.3 required Run aliyun version to verify >= 3.3.3. If not installed or version too low, run curl -fsSL https://aliyuncli.alicdn.com/setup.sh | bash to update. Then [MUST] run aliyun configure set --auto-plugin-install true to enable automatic plugin installation. Then [MUST] run aliyun plugin update to ensure local plugins are up-to-date. At the start of the Core Workflow (before any CLI invocation): [MUST] Enable AI-Mode — AI-mode is required for Agent Skill execution. ``bash aliyun configure ai-mode enable aliyun configure ai-mode set-user-agent --user-agent "AlibabaCloud-Agent-Skills/alibabacloud-sysom-diagnosis" ` At the **end** of the workflow (after success or failure): **[MUST] Disable AI-Mode** — AI-mode must be disabled after skill execution completes. `bash aliyun configure ai-mode disable ``

Core Agent Behaviors

1. Local first: run memory quick checks first; do not ask for region/instance by default.
2. Intent routing:

- Memory domain: high usage/panorama → memgraph; OOM/oom-killer → oom; Java → javamem; unknown → classify. See memory-routing.md. - Non-memory domain: IO/network/load → use corresponding io/net/load subcommands and go directly to remote deep diagnosis. See non-memory-routing.md. - Hard constraint for remote deep diagnosis: for --deep-diagnosis, io/*, net/*, and load/*, SysOM InvokeDiagnosis must be triggered via ./scripts/osops.sh; do not degrade to generic ECS diagnostics APIs or manual Ecs.RunCommand/Cloud Assistant collection (top/ps/iostat) as replacements.

1. Follow envelope instructions: always read agent.summary and execute agent.next. Quick output is signal detection only. If agent.next contains a command, execute it before summarizing to the user.
2. Envelope is source of truth: base conclusions on envelope data; no extra ad-hoc collection is needed.
3. Latency + socket queue backlog: if net netjitter/net packetdrop has been run and looks normal but ss shows high Send-Q/Recv-Q, cross-check with memory memgraph --deep-diagnosis.

For full conventions (execution directory, credential safety, precheck noise reduction, etc.), see agent-conventions.md.

Envelope Output

CLI stdout is a JSON envelope (format: sysom_agent, version: 3.4). The agent directly consumes agent.summary (summary), agent.findings (key metrics), and agent.next (next-step command, executed with Bash in the skill root). Business payload is in data.routing, data.local, and data.remote. See output-format.md.

Precheck / Authentication Failure

On authentication failure, the envelope includes data.remediation (standalone precheck) or data.precheck_gate.remediation (deep-diagnosis merged flow). Guide users according to envelope instructions. See agent-conventions.md.

Subcommand Quick Reference

Memory Domain

IO Domain

Network Domain

Load Domain

Quick Start

cd <sysom-diagnosis>
./scripts/osops.sh memory classify                                          # local classification
./scripts/osops.sh memory memgraph                                          # local memory panorama
./scripts/osops.sh memory memgraph --deep-diagnosis --channel ecs --timeout 300  # remote memory deep diagnosis
./scripts/osops.sh io iofsstat --channel ecs --timeout 300                  # IO dashboard
./scripts/osops.sh net packetdrop --channel ecs --region cn-hangzhou --instance i-xxx  # packet loss diagnosis
./scripts/osops.sh load delay --channel ecs --params '{"duration":30}'      # scheduling delay

For other instances, add --region <id> --instance <i-xxx>. For first-time setup, run ./scripts/init.sh.

Three Remote OpenAPI Requirements

Key Path Index