Investigation playbook

The Adobe Commerce Traffic Insights app is designed to help you investigate the following problems:

  • Bandwidth overage
  • Crawler load
  • Malicious traffic

Alternatively, you can request Advanced Security: native bot management, Layer 7 DDoS & rate limiting, Adobe’s native escalation path when manual mitigation is insufficient. Each step references the widget that surfaces the symptom, so you can move from a metric to a concrete action.

WARNING
The suggestions on this page are only guidelines. Always validate any blocking rule against your own traffic before deploying it.

CDN bandwidth overage

Before considering bandwidth overages, understand how bandwidth is billed. Traffic for all Fastly services bundled with the Adobe Commerce on Cloud Infrastructure account, including every production and staging environment, counts toward the common usage compared against the annual allowance in your contract. Start from Bandwidth > Total Bandwidth, then attribute the volume with Bandwidth By Content Type and Bandwidth By Domain Details.

Media content

Some stores legitimately serve a large share of bandwidth as media because of their catalog. If Bandwidth By Content Type shows a significant amount of media bandwidth, consider the following mitigations:

Large files

Some sites contain large files or specific, heavy responses, for example, Enterprise Resource Planning (ERP) integrations or exports. Use URLs By Bandwidth to review the BW and Avg Size columns to find these large files. You can use Path Segment lvl 1 By Bandwidth for a higher-level view.

Heavy 404s

An Adobe Commerce 404 page not found is usually a heavy, theme-stylized page (~1.5 MB) and non-cacheable, so repeated 404s can generate abnormal traffic. Even a trivial missing resource like favicon.ico can turn into a heavy 404 page instead of a small file. Use the 404 and 404 BW columns in Bandwidth By Domain Details, URLs By Bandwidth, Top IPs By Bandwidth, and Stats By IP Subnets to find clients, IPs, and URLs consistently generating 404 volume. Then reduce or limit that access, for example, return a lightweight 403 instead.

Low FPC hit ratio

Adobe recommends enabling Fastly shielding so a main CDN cache aggregator serves the origin, letting fewer requests reach it from local Points of Presence (POPs) closest to the client. See checking your configuration.

POP-to-client and shield-to-POP traffic are counted separately, and while the client response is compressed, shield-to-POP traffic is not compressed to preserve Edge Side Includes (ESI) support. This means a low Full Page Cache (FPC) hit ratio drives much higher bandwidth on dynamic pages. Confirm the symptom with FPC Hit Ratio, FPC Stats By Domain, and CDN Network Segment Bandwidth.

A low hit rate is often driven by a large volume of search-engine crawlers (see Search bots and crawlers). Another mitigation is to serve a stale cache to crawlers when available. If broad, frequent cache invalidations are the cause, use Cache Invalidation By Tags and FPC Age By Top URLs to find the churned tags/URLs.

Search bots and crawlers

To gauge crawler impact, start in Known Bots By Bandwidth and Known Bots Impact Details to see which bots are most active, then filter by a specific bot to study only its requests.

Too many requests

The most common cause of a search bot sending too many requests occurs while parsing pages that carry <meta name="robots" content="index,follow">. Bots can follow top-navigation and layered-navigation links in a near-endless loop. Consider the following options to address this issue:

WARNING
Consult a Search Engine Optimization (SEO) expert before restraining crawler activity. Retraining can negatively affect your SEO.
  • Add nofollow to top-navigation and layered-navigation links, for example <a rel="nofollow" href="https://example.com/sales.html">Sales</a>.
  • Change the page meta tag to index,nofollow — either as a common design configuration setting or per page type with custom extensions. Keep sitemap.xml accurate so that bots always have an up-to-date list of pages to index.
  • Update robots.txt to block paths and resources bots should not access.
  • Note the crawl-delay directive is not part of the official Robots Exclusion Protocol, but it does work for some bots, such as Bingbot, Slurp, SEMrushBot, and a few others. Googlebot ignores this directive.
  • Add rate-limit rules. There is native abusive crawler protection in the Fastly module. For finer control, a custom Varnish Configuration Language (VCL) snippet can return 429 (Too Many Requests) or 405 (Method Not Allowed) for a user-agent regex with an individual rate limit. Check the crawler’s documentation for the preferred method and response code. See Fastly’s rate-limiting VCL guidance.
  • AI and large language model (LLM) crawlers are a growing special case. They do not always identify themselves consistently, so VCL user-agent rules can lag behind. Adobe’s Advanced Security add-on has native bot management that can distinguish verified from suspected AI crawlers and fetchers at the edge, which VCL alone cannot.

Blocking unwanted crawlers

If certain search engines generate significant traffic and are not important to the business, they can be blocked entirely:

  • Some bots follow robots.txt changes 1–2 days later, after re-reading and updating their parsing rules.
  • If a crawler ignores robots.txt, block it with a custom VCL snippet (example). Some crawlers explicitly document this as the preferred or only method of frequency control.

Malicious scripts & scrapers

Use the Traffic Insights app to identify the common directions of attack, filtering by focus areas as needed. If red-flagged requests come predominantly from certain IPs, subnets, or geolocations (Top IPs By Requests Count, Stats By IP Subnets, Stats By Country), consider blocking them with custom Fastly VCL.

Every Cloud Infrastructure project already has a baseline of automatic protection regardless of any configuration you do. The included Web Application Firewall (WAF) immediately blocks SQL injection and known-malicious-IP signals (backdoor, attack tooling, CMDEXE, Log4J-JNDI, traversal, XSS), and rate-limits other non-malicious IPs once they cross 50 requests/minute, 350 requests/10 minutes, or 1,800 requests/hour. That baseline is what Requests By WAF Response and the WAF signal columns in this app’s tables are indicating. A spike in these columns does not necessarily mean that you are not being protected.

  • Watch for credential stuffing, account takeover, fake-account creation, card testing, content scraping, and inventory/cart hoarding. These bot-driven abuse patterns are surfaced in the Bots Activity and Requests Analysis tab. High-volume, low-diversity traffic hitting login, account, checkout, or catalog endpoints is the signature to look for in Top IPs By Requests Count and Known Bots Impact Details.
  • Protect checkout and checkout API endpoints from bot attacks with Google reCAPTCHA.
  • Use the Fastly module’s native rate-limit path protection.
  • Check Next-Gen WAF signals in the comma-separated Sigsci_Tags field and combine relevant signal matches into a targeted blocking rule. A suspicious request’s value can look like BOT-ANALYSIS,DATACENTER,SIGSCI-IP,SITE-FLAGGED-IP,SUSPECTED-BAD-BOT. The WAF labels an IP with SITE-FLAGGED-IP up to a threshold before it begins blocking automatically. The WAF Attack & Anomaly Signals, WAF Bots Signals, and Requests By WAF Response widgets, and the WAF columns in the IP, subnet, and country tables, surface these.
  • See Adobe’s article on blocking malicious traffic for Adobe Commerce on the Fastly level for common approaches.
  • For complex scenarios where manual blocking is not a viable option, such as sustained bot campaigns, attacks spread across many IPs/APIs, or Layer 7 Distributed Denial of Service (DDoS), consider Adobe’s Advanced Security add-on first (see native bot management). It runs on the same Fastly edge serving your storefront. If you need capabilities outside of its scope, a third-party managed bot-mitigation service with native Fastly integration, such as Datadome or HUMAN Bot Defender (formerly PerimeterX) is the suggested alternative. All of these options add additional costs.

Advanced Security: native bot management, Layer 7 DDoS and rate limiting

The previous sections discuss what can be done with the Traffic Insights app’s data and manual Fastly VCL. For scenarios where that is not enough, such as sustained or evolving bot campaigns, Layer 7 (application-layer) DDoS, or abuse spread thinly across many IPs and API endpoints, Adobe offers Advanced Security.

Advanced Security is a paid add-on for Adobe Commerce on Cloud Infrastructure that adds edge bot management (including AI crawler and fetcher detection), Layer 7 DDoS protection, and advanced rate limiting on the same Fastly platform already serving the storefront. See Advanced Security for full capabilities, current limitations, and how to request it.

Once purchased and enabled, use the Traffic Insights app to verify that Advanced Security is working. Its decisions are reported through the same Sigsci_Tags and Agent_response fields behind WAF Attack & Anomaly Signals, WAF Bots Signals, and Requests By WAF Response. Compare those widgets before and after enabling it to confirm it is actively acting on your traffic.

recommendation-more-help
commerce-on-cloud-help-cloud-guide