skip to content

With Cloudflare's HTTP DDoS Attack Protection, a ticketing site's legitimate on-sale rush was mitigated as a flood; how do you confirm it and fix it without weakening the whole zone?

level: seniorimportance: should knowfreq 34%

answer

  1. find the rule before changing anything
  2. Service equals HTTP DDoS
  3. the generic tag fires on volume
  4. one rule, one step down

basics

~20 s

Filter security events by Service equals HTTP DDoS to find the rule that fired and confirm the traffic was real buyers, then lower only that rule's sensitivity one step at a time, path-scoped only with Advanced DDoS Protection.

solid answer

~50 s

First confirm: in the zone's security analytics, filter events by `Service equals HTTP DDoS`, read the top HTTP DDoS rules and the action taken, and compare the flagged requests with what real buyers look like: your paths, your app's user agent, the countries you sell in, a rate that matches the announced on-sale time. A rule tagged `generic` (request floods) is the likely culprit, since those rules can trigger on unusually high legitimate volume. Then fix narrowly: create an override, browse to that rule by name and lower its sensitivity from High to Medium, then Low only if needed, rather than setting the whole ruleset to *Essentially Off*. On Enterprise use Log to watch it first; with Advanced DDoS Protection scope the override to the on-sale path. It applies within a minute or two. A `read-only` rule cannot be overridden, so that case goes to Cloudflare Support.

go deeper

for a junior

Know where to look: security events filtered by Service equals HTTP DDoS show which rule fired. Know that the fix is an override that lowers sensitivity.

for a middle

Explain why generic flood rules are the usual suspects in a flash crowd, the step-down order High, Medium, Low, Essentially Off, and that Log is an Enterprise option.

for a senior

Fix one rule by one step, scoped by path only with Advanced DDoS Protection. Reject ruleset-wide Essentially Off and user-agent exceptions. Check the flagged traffic for a real attack hiding in the rush.

for a principal

Plan the next on-sale: who reviews DDoS alerts during launches, which rule changes are pre-approved for an event window, and when each temporary override is reverted.

## What happened A ticketing site announces an on-sale at 10:00. At 10:00 thousands of real buyers refresh the event page and hit the seat-selection path at once, a pattern that looks a lot like a layer-7 request flood. The **HTTP DDoS Attack Protection** managed ruleset decides it is one, and buyers see challenges or errors. This is a **false positive**: legitimate traffic classified as attack traffic. The goal is to fix it without opening the zone to the next real attack. ## Step 1: confirm it was the DDoS rules, and which one Do not start tuning until you know which rule fired. In the zone's security analytics: 1. Open the events view and add the filter `Service equals HTTP DDoS`. Other services, such as WAF managed rules or rate limiting rules, are excluded, so you see only DDoS mitigations. Security Events is a Pro-and-above view; a Free zone gets sampled logs only. 2. Scroll to **Top events by source** and the **HTTP DDoS rules** list, and copy the name or ID of the rule that fired. 3. Read the **action** applied. Managed Challenge, Block and Force Connection Close each hurt buyers differently. 4. Check the rule's tag. Rules tagged `generic` mitigate floods with no known signature and carry a higher activation threshold to limit false positives, but they can still fire on unusually high legitimate volume. Rules tagged `botnets` are high-accuracy and should stay enabled. Then compare the flagged requests with known-good traffic: the timing matches the announcement, the paths are your event and checkout pages, the user agent is your mobile app or common browsers, and the source countries are your market. If the flagged traffic instead shows odd headers, one unusual user agent or paths nobody buys from, it may be a real attack timed to the on-sale, and loosening the rule would be the wrong move. A **DDoS alert** speeds this up: the HTTP DDoS Attack Alert, available to WAF/CDN zones, arrives within about a minute and names the matched rule and any override in force. ## Step 2: change as little as possible The narrowest fix depends on the plan: | Plan and subscription | Narrowest fix | |---|---| | Free, Pro, Business | One override: lower **that rule's** sensitivity from High to Medium, then Low or Essentially Off only if it keeps firing | | Enterprise | As above, or set that rule to **Log** to watch it without mitigating | | Enterprise with Advanced DDoS Protection | An override with an expression, for example `starts_with(http.request.uri.path, "/onsale")`, lowering sensitivity only for that path | What to avoid: - **Setting the whole ruleset to Essentially Off.** It fixes the symptom and removes most DDoS protection from every path in the zone. - **Scoping an exception by user agent alone.** `http.user_agent` is a supported override field, but any client can send your app's user agent, so an attacker can copy it. Prefer a path, a host or an IP list for a partner. - **Reaching for a WAF Skip rule.** A custom rule's Skip action can bypass rate limiting rules, Super Bot Fight Mode and WAF managed rules, not the DDoS managed rulesets. To apply the fix, open the **DDoS protection** tab of the zone's security rules, create an override (or edit the existing one if the zone has no override left), browse rules, paste the rule name and pick the new sensitivity or action. It takes effect within one or two minutes, and the analytics should show the change straight away. Rules tagged `read-only` cannot have their sensitivity or action changed. If one of those is firing, contact Cloudflare Support with the rule, the times and the affected paths. ## Step 3: after the event - Review the override once the rush is over. The documented follow-up is to settle the rule at a sensitivity that avoids future false positives and return its action to the default. - Keep the rest of the ruleset at the defaults Cloudflare recommends: High sensitivity and default actions. - Before the next on-sale, Enterprise zones, where Log is available, can switch the relevant rule to Log ahead of time and read what it would have done; others can lower that one rule by one step for the event window. - Make the cache absorb as much of the rush as possible. Event pages served from cache never reach the origin, which also keeps the origin-error rule quiet. ## Why the narrow fix matters A flash crowd and an attack can arrive together; attackers time floods to launches. One rule lowered by one step, ideally scoped to one path, keeps every other detection at full strength.

  • Your app's traffic is still blocked after lowering the rule from High to Medium; what is the documented next step?
    Lower that same rule again, to Low and then Essentially Off if it keeps blocking, still at rule scope. With Advanced DDoS Protection, scope the override to the affected traffic instead of lowering it for everyone. If the rule is tagged read-only, no override can change it and the case goes to Cloudflare Support.
  • How do you tell a flash crowd from a real attack timed to the on-sale, using what the DDoS events show?
    Compare flagged requests with known buyers: paths you sell from, your app's user agent, your market's countries, and a start that matches the announcement. A real attack tends to show traits buyers lack: unusual headers, one odd user agent, paths no buyer requests, sources outside your market. If it looks like an attack, keep the rule and add controls rather than loosening it.

  • Setting the whole HTTP DDoS ruleset to Essentially Off to stop one false positive.
  • Excluding traffic by user agent alone, which attackers can copy.
  • Using a WAF custom rule with Skip to bypass the DDoS managed rulesets.
  • Tuning before checking which rule fired and what action it applied.
  • Expecting an override on a read-only rule to change its behaviour.