huawei-cloud-cdn-abnormal-status-code-analysis

CDN Abnormal Status Code Analysis

Overview

This skill diagnoses abnormal HTTP status codes (4xx / 5xx) on a Huawei Cloud CDN accelerated domain. It is read-only end to end: every action is a query (GET statistics / GET config / GET log-download-link / read-only log fetch). It answers four questions for any 4xx/5xx anomaly:

1. Is there an anomaly, and how big? — quantify 4xx/5xx volume and ratio over a window.
2. Which exact code, when, edge or origin? — drill to the specific code + time bucket, and fork edge-generated vs origin-generated via the back-to-source (bs_*) status statistics.
3. Who / where / what path? — top IPs/paths/UA/referer, client IP count (traffic farming vs real users).
4. Root cause + forensics? — narrow to edge-config rules or origin-side issues, then pin the exact requests via access logs.

Key Features:

• Edge-vs-origin fork driven by one command (bs_status_code_*): result={} ⇒ edge-generated; non-empty ⇒ origin-generated.
• Covers the full 4xx/5xx set: 400/401/403/404/405/406/408/410/412/416/429/499/495/496/500/501/502/503/504/505/530….
• Read-only log helper scripts/fetch_cdn_log.py downloads + decompresses a CDN log and extracts the abnormal-status rows as JSON.
• Structured diagnosis report with a status-code → root-cause decision tree and console/ticket remediation boundary (no CLI write ops).

Tools: hcloud CLI (KooCLI) + Python log helper (scripts/fetch_cdn_log.py). Core Principle: Read-only diagnosis; no configuration changes.

Triggers

Use when the user input matches any of the following (user-phrase level, bilingual):

Language	Trigger examples (user phrasing)
Chinese	"CDN domain 403s suddenly increased, help me check" "Why have there been continuous 502s recently" "4xx/5xx status code anomaly analysis" "Is this issue at the edge or origin? Help distinguish it" "Business exception code troubleshooting" "CDN rate limiting/blocking troubleshooting" "Status code anomaly"
English	"403 spike on my CDN domain" / "root-cause a 5xx burst" / "status code analysis" / "is this edge or origin?" / "abnormal status codes during daily inspection"

Keyword triggers (also in the frontmatter description): status code anomaly, business exception code, 4xx, 5xx, 403, 404, 502, 503, 504, status code, abnormal status, CDN anomaly, edge/origin, rate limiting, status code analysis, edge vs origin.

Near-miss / Do NOT use

Do NOT use this skill when:

Scenario	What to do instead
The domain is not onboarded to Huawei Cloud CDN	Confirm CDN onboarding / domain existence first; non-CDN HTTP issues are out of scope
The user wants to change CDN config / refresh / preheat / ban URLs / enable-disable domains	Refuse — this skill is read-only; guide to the CDN console or manual hcloud CLI (see Prohibited Operations)
Pure bandwidth / traffic analysis with no status-code focus	Not this skill's scope (no 4xx/5xx anomaly to explain)
User pastes AK/SK or asks to configure credentials in chat	Refuse — see Authentication; emit the secure setup template and wait
Non-CDN business exceptions (ECS / ELB / OBS / application-layer issues unrelated to CDN semantics)	Redirect to the relevant service skill

⛔ Prohibited Operations (Security Constraints)

This skill strictly prohibits all non-GET (write/modify/delete) CDN operations, even if requested by the user.

Total: 55 prohibited operations (24 POST + 25 PUT + 6 DELETE).

For the complete list of all 55 prohibited non-GET operations with risk descriptions, see references/prohibited-operations.md.

Representative prohibited operations (full list in the reference doc):

Prohibited Operation	Example CLI	Reason
❌ Create domain	CreateDomain (v1/v2), CreateDomainByDuplicate	Write op, creates production resource
❌ Delete domain	DeleteDomain (v1/v2)	Irreversible
❌ Modify domain config	hcloud CDN UpdateDomainFullConfig / UpdateBlackWhiteList / UpdateRefer / UpdateResponseHeader / UpdateOriginHost / UpdateCacheRules / UpdateHttpsInfo	Write op, affects production traffic
❌ Enable / disable domain	EnableDomain (v1/v2), DisableDomain (v1/v2)	Write op, affects production traffic
❌ Refresh / preheat	hcloud CDN CreateRefreshTasks / CreatePreheatingTasks	Write op, changes cache
❌ Modify billing mode	SetChargeModes	Financial impact
❌ Verify domain owner	hcloud CDN VerifyDomainOwner	Write op
❌ Set stats config / export open	SetStatsConfig / ExportStatsOpen	Write op

If the user requests a prohibited operation, refuse and inform: "Per security constraints, this skill performs read-only diagnosis only and does not run any write/delete operations. To change CDN configuration, use the Huawei Cloud CDN console or run the relevant hcloud command manually — this skill will not execute it. The complete list of 55 prohibited operations is documented in references/prohibited-operations.md."

Architecture

CDN Abnormal Status Code Analysis (read-only)
├── ListDomains/v2                       (target domain → domain_id)
├── ShowDomainStats/v2
│   ├── summary  http_code_2xx..5xx + req_num    → quantify 4xx/5xx volume + ratio
│   └── detail   status_code_4xx/5xx             → exact code + time array
├── ShowDomainStats/v2  bs_status_code_4xx/5xx    → ★ edge vs origin fork ★
├── ShowBandwidthCalc / ListDomainClientStats      → traffic-spike / 刷量 correlation
├── Top-N family                                    → who / where / what path
├── Root cause fork
│   ├── origin side: ShowOriginHost / ShowDomainDetail / ShowHistoryTasks / TopOriginUrl
│   └── edge side:   ShowDomainFullConfig / ShowRefer / ShowBlackWhiteList /
│                    ListRuleDetails / ListBanUrl / ListAccessControlTask /
│                    ShowResponseHeader / ShowCertificatesHttpsInfo
├── ShowLogs/v2 + python scripts/fetch_cdn_log.py  → per-request forensics (JSON)
├── ShowIpInfo/v2                                   → client/edge IP attribution
└── generate diagnosis report

API Call Budget

Step	API / Script	Rate Limit	Est. Time
1	hcloud CDN ListDomains/v2 --cli-region={region}	—	<2s
2	hcloud CDN ShowDomainStats/v2 --cli-region={region} (summary + detail + bs)	5/s	<3s
3	Top-N family + ListDomainClientStats	2-5/s	<5s
4	hcloud CDN ShowDomainFullConfig/v2 --cli-region={region} + Show* config	—	<5s
5	hcloud CDN ShowLogs/v2 --cli-region={region} + python scripts/fetch_cdn_log.py	—	≤ 30s

Total estimated time: < 45 seconds (log-heavy window may be longer).

KooCLI Command Format Standard

hcloud CDN <Operation> --cli-region=<region> [--key=value ...]

• Service name: CDN (uppercase). Operation name: PascalCase (e.g., ShowDomainStats/v2).
• Region parameter: --cli-region=<region>. CDN supports cn-north-1 (Beijing) and ap-southeast-1 (Singapore); results are region-independent. Recommended: cn-north-1 uniformly. (cn-north-4 is NOT supported by CDN.)
• Parameter format: --key=value (equals sign, no spaces). Space-separated form is unsupported.
• Output: append --cli-output=json for machine parsing.

Prerequisites

Pre-check: hcloud (KooCLI) >= 3.2.0 — hcloud version. If missing, see references/cli-installation-guide.md.

Pre-check: Python >= 3.8 — python --version. Required by scripts/fetch_cdn_log.py.

Pre-check: Python requests >= 2.25 — python -c "import requests; print('ok')". The log helper downloads via requests. Install if missing: pip install requests>=2.25

Pre-check: hcloud credentials — hcloud configure list. If invalid, stop and guide secure configuration. Never read/echo/print AK/SK; never ask the user to paste AK/SK in chat.

⚠️ hcloud parameter format: all params must use --key=value.

⚠️ CDN region: use cn-north-1 uniformly (CDN does not support cn-north-4).

⚠️ Time / interval rules: time = ms timestamp, [start,end) left-closed right-open; interval=300 (≤2d, 5-min points) / 3600 (≤7d, hour points) / 86400 (≤31d, CST 0:00 points). status_code_* (edge) and bs_status_code_* (origin) cannot be mixed in one query.

Authentication

• Prohibited: reading/echoing/printing AK/SK; asking users to input AK/SK in chat; hcloud configure set with plaintext; accepting AK/SK pasted by the user.
• Allowed: read credentials from environment variables or the active CLI profile only.
• If the user pastes AK/SK, stop, refuse, and emit the secure setup template (export HUAWEICLOUD_SDK_AK=... / HUAWEICLOUD_SDK_SK=...), then wait.
• Verify: hcloud configure list — confirm a valid profile exists; if not, stop.

IAM Permission Policies

See references/iam-policies.md. Minimum: cdn:*:query* (plus cdn:configuration:queryDomains) and cdn:log:* for log forensics. This skill is strictly read-only: no create / update / delete / refresh / ban permission is required or included.

Core Commands

Command	Purpose	Key Parameters
hcloud configure list	Check credentials	none
hcloud CDN ListDomains/v2 --cli-region=<region>	List domains → domain_id	—
hcloud CDN ShowDomainStats/v2 --cli-region=<region> --action=summary|detail --domain_name=<d> --stat_type=<t> --start_time=<ms> --end_time=<ms> --interval=<s>	Quantify / localize status codes	--action, --stat_type, --interval
hcloud CDN ShowBandwidthCalc --cli-region=<region> --domain_name=<d> --calc_type=bw_peak --start_time=<ms> --end_time=<ms>	Bandwidth peak (traffic-spike correlation)	--calc_type
hcloud CDN ListDomainClientStats --cli-region=<region> --domain_name=<d> --stat_type=ip_num --start_time=<ms> --end_time=<ms>	Client IP count (traffic farming vs real users)	--stat_type=ip_num
hcloud CDN ListCdnDomainTopIps --cli-region=<region> --domain_name=<d> --stat_type=req_num --start_time=<ms> --end_time=<ms>	Top-N drill (TopIps; same form for Path/OriginUrl/Refers/Uas; req_num only, no status)	--stat_type=req_num
hcloud CDN ShowDomainFullConfig/v2 --cli-region=<region> --domain_name=<d>	Full edge config (auth/refer/ip_filter/ua/…)	—
hcloud CDN ShowRefer --cli-region=<region> --domain_id=<id>	Referer anti-hotlinking config	--domain_id
hcloud CDN ShowBlackWhiteList --cli-region=<region> --domain_id=<id>	IP blacklist/whitelist	--domain_id
hcloud CDN ListRuleDetails --cli-region=<region> --domain_name=<d>	Rule engine block rules	—
hcloud CDN ListBanUrl --cli-region=<region> --start_time=<ms> --end_time=<ms> --page_number=1 --page_size=50	Banned URLs (may need ticket whitelist)	time range
hcloud CDN ListAccessControlTask --cli-region=<region> --start_time=<ms> --end_time=<ms> --limit=100	Ban/unban tasks (may need ticket whitelist)	time range
hcloud CDN ShowOriginHost --cli-region=<region> --domain_id=<id>	Origin HOST	--domain_id
hcloud CDN ShowResponseHeader --cli-region=<region> --domain_id=<id>	Response header / error-page config	--domain_id
hcloud CDN ShowDomainDetail --cli-region=<region> --domain_id=<id>	Domain detail (origin/status/CNAME)	--domain_id
hcloud CDN ShowHistoryTasks/v2 --cli-region=<region> --file_type=file --page_number=1 --page_size=50	Refresh/preheat history	—
hcloud CDN ShowCertificatesHttpsInfo/v2 --cli-region=<region> --domain_name=<d>	HTTPS cert (49x/TLS)	--domain_name
hcloud CDN ShowLogs/v2 --cli-region=<region> --domain_name=<d> --start_time=<ms> --end_time=<ms> --page_size=100	Log download links (single domain, ≤30 days)	--domain_name
python scripts/fetch_cdn_log.py --url <link> [--status 403,502,503,504] [--timeout 30] [--max-lines 200]	Fetch+decompress log, extract abnormal rows (emits JSON)	--url, --status
hcloud CDN ShowIpInfo/v2 --cli-region=<region> --ips=<ip1>,<ip2>	IP attribution (≤20)	--ips

Top-N family stat_type supports only flux/req_num (no status code) — cross-reference with ShowDomainStats/logs.

Parameter Confirmation

Parameter	Required	Description	Default	Example
domain_name	Yes	CDN accelerated domain to diagnose	—	www.example.com
domain_id	Yes (for some Show* by-id ops)	From ListDomains/v2	—	d3aaa521…
--cli-region	Yes	CDN region	cn-north-1	cn-north-1
time window	Yes	[start,end) ms timestamps, aligned to interval points	—	last 7 days
--status (log helper)	No	Status codes to extract	403,404,499,500,502,503,504,530	403,502

User confirmation checklist:

• Target domain provided
• User understands this is read-only diagnosis
• User understands log fetch has a 30s timeout

Core Workflows

Target domain is required before any diagnosis step.

• If the user did not provide a target domain, ask the user for the domain name and wait for the reply before starting.
• Only if the user does not know the domain or asks you to look it up, list the account's domains via hcloud CDN ListDomains/v2 and ask the user to choose one.
• Never start the diagnosis without an explicit user-provided domain: do not guess a domain, do not fall back to an example/default domain, and do not pick a domain from the list yourself.

Step 1: Discovery & Quantification — find 4xx/5xx, size, traffic correlation

📄 Detailed steps → references/task-discovery.md

Step 2: Localize & Fork — exact code + time + edge/origin

📄 Detailed steps → references/task-localize.md

Step 3: Distribution — top-N + client-IP count

📄 Detailed steps → references/task-distribution.md

Step 4: Root Cause — origin side (4A) / edge side (4B) per Step 2 fork

📄 Detailed steps → references/task-rootcause.md

Step 5: Per-request Forensics — log fetch + IP attribution

📄 Detailed steps → references/task-forensics.md

Step 6: Report — decision tree + remediation boundary

📄 Detailed steps → references/task-report.md

FAQ (Q&A)

Detailed troubleshooting and the full error-code reference are in references/troubleshooting.md.

Q1: Why do ListBanUrl / ListAccessControlTask return CDN.0004?

A: These two query APIs require a ticket (support-ticket) whitelist; they are still query-class GET operations. Record the error, do not bypass, continue with the other evidence, and mark "unconfirmed via CLI" in the report.

Q2: Why does ShowDomainStats return an empty result?

A: No traffic in the window, the status code did not occur, or the timestamps are not aligned to the interval (interval=86400 requires CST 0:00 points). Widen the window (e.g. 31 days), align the time points, and confirm the domain had traffic, then retry.

Q3: Can status_code_* and bs_status_code_* be combined in one query?

A: No. Edge and back-to-source statistics must be run as two separate ShowDomainStats/v2 calls.

Q4: Why is there no status code in the Top-N results?

A: The Top-N family (TopIps/Path/OriginUrl/Refers/Uas) stat_type supports only flux/req_num — there is no status-code dimension. Use Top-N to shortlist suspect IPs/paths, then cross-check the actual status codes via ShowLogs / ShowDomainStats.

Q5: Why does ShowLogs return logs: [] / total: 0?

A: No log file covers the window, the window is not left-closed right-open / hour-aligned, or the domain had no traffic in that period. Align to whole-hour CST points, widen the window, and confirm traffic, then retry.

Q6: How do I handle fetch_cdn_log.py errors?

A: See the error-code table in references/troubleshooting.md (9 error.reason values: connect_timeout / connect_failed / http_error / bad_gzip / invalid_url / invalid_timeout / invalid_max_lines / invalid_status / missing_library). Soft failures exit 0 (JSON error included), argument/missing-library errors exit 2. Common fixes: raise --timeout (≤60), re-run ShowLogs/v2 to refresh an expired link, or python -m pip install requests>=2.25.

Q7: Why does a CDN query return 403 Insufficient permission?

A: The IAM user lacks the CDN read-only scope. Grant cdn:*:query* (plus cdn:configuration:queryDomains and cdn:log:* for log forensics) per references/iam-policies.md. Never elevate to any write action.

Known Issues and Limitations

Limitation	Description
ShowLogs/v2	Single domain only; span ≤ 30 days
Top-N stat_type	Only flux/req_num; no status-code dimension (cross-check required)
ListDomainClientStats	ip_num supports 1-day granularity only (CST 0:00); the unique-visitor metric supports 5-min granularity
ListBanUrl / ListAccessControlTask	Require a ticket (support-ticket) whitelist (CDN.0004); record and continue
Log parsing	fetch_cdn_log.py parses the official Huawei Cloud 14-field log format; fields containing spaces (time, user_agent) are kept as one token / quoted (official doc)
CDN region	Only cn-north-1 / ap-southeast-1; cn-north-1 recommended uniformly; cn-north-4 is not supported

Error Code Quick Reference

Error	Meaning	Action
CDN.0004	Not in the ticket whitelist (ListBanUrl / ListAccessControlTask)	Record, continue, mark "unconfirmed via CLI"
cli-region的值不支持	Region not supported by CDN	Use --cli-region=cn-north-1 (or ap-southeast-1)
403 Insufficient permission	IAM lacks the CDN read-only scope	Grant cdn:*:query* (+ cdn:log:*); never elevate

Failure Modes

Detailed troubleshooting is in references/troubleshooting.md. Each task step's Exit criteria is the self-check.

Known failure modes and recovery strategy

Failure mode	Symptom	Recovery strategy (retry / fallback / degrade / report)
Region not supported	cli-region的值不支持, only cn-north-4/ru-moscow-1 hinted	Switch to --cli-region=cn-north-1 (or ap-southeast-1)
CDN.0004 whitelist	ListBanUrl / ListAccessControlTask return "not in the whitelist"	Degrade: record unconfirmed via CLI, continue with other evidence, mark it in the report; do not bypass
Empty result	ShowDomainStats returns {}	Retry: widen the window (31 days), align to the interval points (CST 0:00 for 86400), confirm traffic
Edge/origin mixed query	Passing status_code_* and bs_status_code_* together errors	Split into two ShowDomainStats/v2 calls
Top-N has no status codes	Top-N returns request counts only	Cross-check: confirm the actual codes via ShowLogs / ShowDomainStats
ShowLogs returns no links	logs: [] / total: 0	Retry: align to whole-hour CST points, widen the window, confirm traffic in the window
Log download failure	fetch_cdn_log.py error.reason = connect_timeout / connect_failed / http_error / bad_gzip	Retry / refresh: raise --timeout (≤60), re-run ShowLogs/v2 to refresh the expired link, check gzip
Permission denied	Query returns 403 Insufficient permission	Check the IAM read-only scope (cdn:*:query* + cdn:log:*); never elevate to write actions

Error propagation and handling protocol (fetch_cdn_log.py)

• exit 0: probe ran to completion (including soft failures — HTTP non-200, empty result, bad gzip, timeout); emits JSON error.reason for the caller.
• exit 2: argument error or missing dependency library (requests).
• error.reason enum (9 values): connect_timeout / connect_failed / http_error / bad_gzip / invalid_url / invalid_timeout / invalid_max_lines / invalid_status / missing_library.

Self-check before output

• Does the report contain all required sections (Summary / Distribution / Root Cause / Conclusion / Remediation boundary)?
• Does every claim have evidence (config state / bs verdict / log rows / rule list)?
• Are unconfirmed items (e.g. blocked by CDN.0004) marked "unconfirmed via CLI"?
• Does the report avoid recommending any write operation to be executed by this skill? (must be yes)

Output Format

This skill has two output contracts.

1. Diagnosis report (plain text)

Report template: references/task-report.md. Required sections and fields:

Section	Required fields
Header	Analysis Time (ISO8601), Target Domain, Region
Summary	Abnormal status-code set, anomaly window (CST), volume / ratio, edge-origin verdict (bs_status_code_<X> empty → edge / non-empty → origin)
Distribution	Top IP/URL/UA/Referer, client verdict (few-IP burst / many real users, with ip_num/req_num), bandwidth peak
Root Cause	Candidate mechanism, evidence, ruled-out items, unconfirmed-via-CLI items
Conclusion	Status (Edge-generated / Origin-generated / Partial), suggestion, remediation boundary (write ops only via the CDN console or manual hcloud CLI)

2. fetch_cdn_log.py JSON (structured)

Schema: references/related-apis.md. Top-level fields:

Field	Type	Description
url	string	Log download link
http_status	int	Download HTTP status
byte_size	int	Downloaded bytes
status_filter	int[]	Status-code filter set
rows[]	object[]	Matched rows (field enum: status(int 100-599), cache_status(HIT/MISS), time, ...), incl. client_ip, url, user_agent, edge_node_ip
count	int	Number of matched rows
duration_ms	int	Elapsed time
error	object/null	null or {reason, message} (reason enum — see Failure Modes)

Output example (fetch_cdn_log.py JSON excerpt)

{"status":403,"client_ip":"120.46.140.45","url":"/index.html","cache_status":"HIT","user_agent":"Mozilla/5.0 (Windows NT 10.0; Win64; x64)","edge_node_ip":"39.136.130.12","time":"[10/Aug/2026:15:55:13 +0800]"}

Content to NEVER include in the output

• AK/SK and any credentials (see Authentication)
• Internal / sensitive addresses and customer private information (output only the fields needed for diagnosis)
• The raw full log text (output only the extracted rows fields)
• Conclusions beyond the evidence (unconfirmed items must be marked "unconfirmed")

References

Document	Description
task-discovery.md	Step 1: discovery & quantification
task-localize.md	Step 2: localize exact code + edge/origin fork
task-distribution.md	Step 3: top-N/client distribution
task-rootcause.md	Step 4: origin-side (4A) / edge-side (4B) root cause
task-forensics.md	Step 5: per-request log forensics
task-report.md	Step 6: diagnosis report + decision tree
prohibited-operations.md	All 55 prohibited non-GET operations (POST/PUT/DELETE)
dataflow-diagram.md	Mermaid data flow diagram
related-apis.md	API/CLI reference + stat_type + status-code map
iam-policies.md	IAM least-privilege policies
verification-method.md	Verification method
acceptance-criteria.md	Acceptance criteria checklist
troubleshooting.md	Troubleshooting
cli-installation-guide.md	CLI installation guide

Changelog / Version History

Version	Date	Maintainer	Changes
v1.0.0	2026-08-10	cdn-ops	Initial release: read-only 6-step diagnosis workflow (discovery / localize / distribution / root-cause / forensics / report), 55 prohibited operations, fetch_cdn_log.py log-forensics helper, 14 reference documents

Versioning follows SemVer (major.minor.patch); current frontmatter version: 1.0.0. Append a row to the table and bump the frontmatter version on every major change. Maintainer: cdn-ops. If a CDN query API this skill depends on (e.g. ShowLogs/v2, bs_status_code_* statistics) is deprecated / sunset or changes behavior, update references/related-apis.md and references/troubleshooting.md accordingly.